Spaces:
Running
Running
| <html lang="en"><head><meta charset="utf-8"> | |
| <title>VTWM ablation — short 16-frame rollouts (auto-uploaded)</title> | |
| <style> | |
| body{font-family:system-ui,sans-serif;margin:24px;max-width:1400px;color:#222;background:#fafafa} | |
| h1{margin-top:0} | |
| h2{margin-top:40px;padding-bottom:6px} | |
| h2.cat-header{margin-top:48px;border-top:3px solid #333;padding-top:12px;background:#f0f4f8;padding-left:8px;font-size:22px} | |
| h3{margin-top:24px;border-bottom:1px solid #ccc;padding-bottom:4px;font-size:17px} | |
| nav{padding:8px 0;border-bottom:1px solid #ddd;z-index:10} | |
| nav.top{position:sticky;top:0;background:#fafafa;font-size:14px} | |
| nav a{margin-right:8px;text-decoration:none;color:#0066cc} | |
| table{border-collapse:collapse;margin:12px 0 24px 0;background:white;font-size:14px} | |
| th,td{border:1px solid #ddd;padding:6px 10px;text-align:right} | |
| th:first-child,td:first-child{text-align:left} | |
| table.setup th{background:#f0f4f8} | |
| .grid{display:grid;grid-template-columns:repeat(auto-fit,minmax(320px,1fr));gap:12px;margin-top:8px} | |
| figure{margin:0} figcaption{font-size:12px;color:#555;margin-bottom:4px} | |
| video{display:block;width:100%;background:#000;border-radius:4px} | |
| details.sample{margin:8px 0;padding:8px;background:white;border:1px solid #ddd;border-radius:6px} | |
| details.sample summary{cursor:pointer;font-size:14px;padding:4px 0;color:#0066cc} | |
| details.sample[open] summary{color:#222;border-bottom:1px dashed #eee;margin-bottom:8px} | |
| .panel img{display:block;margin:8px 0} | |
| .samples{margin:8px 0 32px 8px} | |
| .badge{background:#ffefc1;color:#7a5400;padding:2px 8px;border-radius:10px;font-size:11px;font-weight:600} | |
| .badge.sota{background:#c6f6c8;color:#0a5b0b} | |
| .badge.blowup{background:#fde0e0;color:#8a1a1a} | |
| .run-desc{color:#555;font-size:13px;margin:4px 0 12px 0;max-width:900px} | |
| .tldr{background:#fffbe6;border-left:4px solid #f0b400;padding:14px 18px;margin:16px 0;border-radius:4px;line-height:1.6} | |
| .tldr h3{margin:0 0 6px 0;border:0;font-size:16px;color:#7a5400} | |
| </style></head><body> | |
| <h1>VTWM ablation — short 16-frame rollouts (auto-uploaded)</h1> | |
| <p style="color:#555">Diffusion-forcing world model rollouts. Each sample is ~2.7 s (16 raw frames @ 6 fps). | |
| Side-by-side videos show <b>ground truth on the left</b>, <b>model prediction on the right</b>. | |
| First half of each clip = history (GT), second half = autoregressively sampled future. | |
| Click a sample header to expand its videos (default-collapsed to keep the page light).</p> | |
| <div class="tldr"><h3>Short 16-frame rollouts (fixed-window inference)</h3>Model predicts the future 8 frames given the past 8 as context. Long-horizon 40s AR rollouts live on the sister page: <a href="https://huggingface.co/spaces/yxma/vtwm-long-horizon-40s">yxma/vtwm-long-horizon-40s</a>.</div> | |
| <nav class="top"><a href="#cat_B._episode_p01_ref_(rollout-consistent)"><b>B. episode_p01 ref (rollout-consistent)</b></a> · <a href="#cat_C._Architecture_—_MoT___hybrid"><b>C. Architecture — MoT / hybrid</b></a> · <a href="#cat_D._Single-modality_controls"><b>D. Single-modality controls</b></a> · <a href="#cat_E._Cosmos_tokenizer"><b>E. Cosmos tokenizer</b></a></nav> | |
| <h2>Sections</h2> | |
| <h2>Setup comparison</h2><p>Click a run name to jump to its rollout videos. ✓ = feature on, ✗ = off. Lower val_loss is better.</p><table class="setup"><tr><th>run</th><th>fusion</th><th>views</th><th>tactile</th><th>shift16</th><th>delta-ref</th><th>cam-pose</th><th>gate</th><th>val_loss</th></tr><tr><td><a href="#group_10_mv1_v2_stride2_p01_val_0.0159"><b>stride2 + p01</b></a></td><td>mmdit_mv</td><td>3</td><td>2</td><td><span style="color:#0a0">✓</span></td><td>episode_p01</td><td><span style="color:#a00">✗</span></td><td>—</td><td>0.0159</td></tr><tr><td><a href="#group_10r_mv1_v2_stride2_p01rand_val_0.0140"><b>p01 + random no-contact ref</b></a></td><td>mmdit_mv</td><td>3</td><td>2</td><td><span style="color:#0a0">✓</span></td><td>episode_rand</td><td><span style="color:#a00">✗</span></td><td>—</td><td>0.014</td></tr><tr><td><a href="#group_11_mot_vt_p01_val_0.0165"><b>MoT V+T</b></a></td><td>mot_vt</td><td>3</td><td>2</td><td><span style="color:#0a0">✓</span></td><td>episode_p01</td><td><span style="color:#a00">✗</span></td><td>—</td><td>0.0165</td></tr><tr><td><a href="#group_11b_mot_vta_p01_val_0.0167"><b>MoT V+T+Action</b></a></td><td>mot_vta</td><td>3</td><td>2</td><td><span style="color:#0a0">✓</span></td><td>episode_p01</td><td>—</td><td>—</td><td>0.0167</td></tr><tr><td><a href="#group_12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156"><b>stride2 + p01 + hybrid + RoPE</b></a></td><td>mmdit_mv_hybrid</td><td>3</td><td>2</td><td><span style="color:#0a0">✓</span></td><td>episode_p01</td><td>—</td><td>—</td><td>0.0156</td></tr><tr><td><a href="#group_12hb_mv1_v2_stride2_p01_hybrid_val_0.0161"><b>stride2 + p01 + hybrid (dual6+single7)</b></a></td><td>mmdit_mv_hybrid</td><td>3</td><td>2</td><td><span style="color:#0a0">✓</span></td><td>episode_p01</td><td>—</td><td>—</td><td>0.0161</td></tr><tr><td><a href="#group_12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159"><b>stride2 + p01 + hybrid (deeper, depth14)</b></a></td><td>mmdit_mv_hybrid</td><td>3</td><td>2</td><td><span style="color:#0a0">✓</span></td><td>episode_p01</td><td>—</td><td>—</td><td>0.0159</td></tr><tr><td><a href="#group_12v_mv1_v2_stride2_p01_vision_only_val_0.0093"><b>vision-only control</b></a></td><td>mmdit_mv</td><td>3</td><td>0</td><td>—</td><td>—</td><td><span style="color:#a00">✗</span></td><td>—</td><td>0.0093</td></tr><tr><td><a href="#group_13t_mv1_v2_stride2_p01_tactile_only_val_0.0062"><b>tactile-only control</b></a></td><td>mmdit_mv</td><td>0</td><td>2</td><td><span style="color:#0a0">✓</span></td><td>episode_p01</td><td><span style="color:#a00">✗</span></td><td>—</td><td>0.0062</td></tr><tr><td><a href="#group_20c_mv1_v2_stride2_p01_cosmos_v3_causal_val_0.1647"><b>Cosmos v3 + causal backbone</b></a></td><td>mmdit_mv</td><td>3</td><td>2</td><td>—</td><td>episode_p01</td><td>—</td><td>—</td><td>0.1647</td></tr><tr><td><a href="#group_20d_mv1_v2_stride2_p01_cosmos_v3_contact_val_0.1301"><b>Cosmos v3 + contact-aware tactile loss</b></a></td><td>mmdit_mv</td><td>3</td><td>2</td><td>—</td><td>episode_p01</td><td>—</td><td>—</td><td>0.1301</td></tr></table> | |
| <h2>Rollout-latent metrics (10 val samples)</h2><p>Latent-space MSE of predicted future vs ground truth. Contact/no-contact split uses tactile latent-to-reference energy (threshold 0.05).</p><table><tr><th>run</th><th>view MSE</th><th>TL MSE</th><th>TR MSE</th><th>tactile contact MSE</th><th>tactile no-contact MSE</th></tr><tr><td><b>stride2 + p01</b></td><td>n/a</td><td>n/a</td><td>n/a</td><td>n/a</td><td>n/a</td></tr><tr><td><b>p01 + random no-contact ref</b></td><td>n/a</td><td>n/a</td><td>n/a</td><td>n/a</td><td>n/a</td></tr><tr><td><b>MoT V+T</b></td><td>n/a</td><td>n/a</td><td>n/a</td><td>n/a</td><td>n/a</td></tr><tr><td><b>MoT V+T+Action</b></td><td>n/a</td><td>n/a</td><td>n/a</td><td>n/a</td><td>n/a</td></tr><tr><td><b>stride2 + p01 + hybrid + RoPE</b></td><td>n/a</td><td>n/a</td><td>n/a</td><td>n/a</td><td>n/a</td></tr><tr><td><b>stride2 + p01 + hybrid (dual6+single7)</b></td><td>n/a</td><td>n/a</td><td>n/a</td><td>n/a</td><td>n/a</td></tr><tr><td><b>stride2 + p01 + hybrid (deeper, depth14)</b></td><td>n/a</td><td>n/a</td><td>n/a</td><td>n/a</td><td>n/a</td></tr><tr><td><b>vision-only control</b></td><td>n/a</td><td>n/a</td><td>n/a</td><td>n/a</td><td>n/a</td></tr><tr><td><b>tactile-only control</b></td><td>n/a</td><td>n/a</td><td>n/a</td><td>n/a</td><td>n/a</td></tr><tr><td><b>Cosmos v3 + causal backbone</b></td><td>n/a</td><td>n/a</td><td>n/a</td><td>n/a</td><td>n/a</td></tr><tr><td><b>Cosmos v3 + contact-aware tactile loss</b></td><td>n/a</td><td>n/a</td><td>n/a</td><td>n/a</td><td>n/a</td></tr></table> | |
| <h2 id="analysis">Training loss curves & conclusions</h2> | |
| <p style="margin:4px 0 12px">📈 <b>Live interactive curves:</b> <a href="https://wandb.ai/yuxiangma/tactile-world-model-diffusion" target="_blank" rel="noopener">https://wandb.ai/yuxiangma/tactile-world-model-diffusion</a> <span style="color:#888;font-size:12px">(wandb — always current; static PNG below is a snapshot)</span></p> | |
| <figure><figcaption>val/train loss vs epoch (visual / tactile / combined / train-visual)</figcaption><a href="analysis/loss_curves_summary.png" target="_blank"><img src="analysis/loss_curves_summary.png" style="max-width:100%;border:1px solid #ddd"/></a></figure> | |
| <details open><summary><b>Conclusions (markdown)</b></summary><pre style="background:#fff;border:1px solid #ddd;padding:12px;white-space:pre-wrap"># Training loss curves — conclusions | |
| Regenerated from wandb (`scripts/plot_loss_curves.py`). Runs covered: **37**. | |
| - **shift16 + delta-to-ref** are the two tactile fixes; without shift16 the tactile stream diverges. | |
| - **cam-pose** curves look fine on val_loss but decoded LPIPS is far worse — see perceptual table. | |
| - **stride2 (4x data)** is the biggest perceptual win; p01 trades a little val_loss for rollout consistency. | |
| - **hybrid_flow** uses rectified flow — its val_loss is on a different scale, judge by decoded metrics. | |
| Included runs: mm_v0, mot_vt_p01, mot_vta_p01, mv1_256_delta_shift, mv1_3v2t_delta_ref, mv1_p01, mv1_p3_delta_shift, mv1_v2_perframe, mv1_v2_stride2, mv1_v2_stride2_p01_adaptw, mv1_v2_stride2_p01_hybrid, mv1_v2_stride2_p01_hybrid_deep, mv1_v2_stride2_p01_hybrid_flow, mv1_v2_stride2_p01_hybrid_rope, mv1_v2_stride2_p01_plucker, mv1_v2_stride2_p01_tactile_only, mv1_v2_stride2_p01_v2t, mv1_v2_stride2_p01_vision_only, mv1_v2_stride2_p01rand, mv2_3v2t_campose_delta_ref, mv2_p4_delta_shift, p2_mv, p3_mv2t, p4_gate, p5_gate, vo_left, vo_middle, vo_right, vo_v0, vtwm_e2_delta_ref, vtwm_e3_delta_ref_null, vtwm_n1_shift16_delta_ref_null_prenoise, vtwm_n2_shift16_delta_ref_tactile_dropout, vtwm_p0_shift16_delta_episode_p01_ref, vtwm_s1_shift16_mmdit_mv, vtwm_s2_shift16_delta_ref, vtwm_s3_shift16_delta_ref_weight | |
| _Not on wandb yet (curve omitted): mv1_v2_h4f12, mv1_v2_stride2_p01, p4_mv2t_extr, vtwm_e1_1v2t_mmdit_mv_ | |
| </pre></details> | |
| <h2 id="cat_B._episode_p01_ref_(rollout-consistent)" class="cat-header">B. episode_p01 ref (rollout-consistent)</h2> | |
| <h3 id="group_10_mv1_v2_stride2_p01_val_0.0159">stride2 + p01</h3> | |
| <p class="run-desc">Stride2 with tactile_ref_source=episode_p01. Best perceptual quality overall (view LPIPS 0.075) despite higher latent val_loss than window. | |
| Showing 8 of 10 samples.</p> | |
| <div class="samples"> | |
| <details id="10_mv1_v2_stride2_p01_val_0.0159_sample_000" class="sample"><summary><b>sample_000</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="10_mv1_v2_stride2_p01_val_0.0159/sample_000/panel.png" target="_blank"><img loading="lazy" src="10_mv1_v2_stride2_p01_val_0.0159/sample_000/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 10_mv1_v2_stride2_p01_val_0.0159/sample_000"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="10_mv1_v2_stride2_p01_val_0.0159/sample_000/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="10_mv1_v2_stride2_p01_val_0.0159/sample_000/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="10_mv1_v2_stride2_p01_val_0.0159/sample_000/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="10_mv1_v2_stride2_p01_val_0.0159/sample_000/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="10_mv1_v2_stride2_p01_val_0.0159/sample_000/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="10_mv1_v2_stride2_p01_val_0.0159_sample_001" class="sample"><summary><b>sample_001</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="10_mv1_v2_stride2_p01_val_0.0159/sample_001/panel.png" target="_blank"><img loading="lazy" src="10_mv1_v2_stride2_p01_val_0.0159/sample_001/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 10_mv1_v2_stride2_p01_val_0.0159/sample_001"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="10_mv1_v2_stride2_p01_val_0.0159/sample_001/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="10_mv1_v2_stride2_p01_val_0.0159/sample_001/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="10_mv1_v2_stride2_p01_val_0.0159/sample_001/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="10_mv1_v2_stride2_p01_val_0.0159/sample_001/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="10_mv1_v2_stride2_p01_val_0.0159/sample_001/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="10_mv1_v2_stride2_p01_val_0.0159_sample_002" class="sample"><summary><b>sample_002</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="10_mv1_v2_stride2_p01_val_0.0159/sample_002/panel.png" target="_blank"><img loading="lazy" src="10_mv1_v2_stride2_p01_val_0.0159/sample_002/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 10_mv1_v2_stride2_p01_val_0.0159/sample_002"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="10_mv1_v2_stride2_p01_val_0.0159/sample_002/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="10_mv1_v2_stride2_p01_val_0.0159/sample_002/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="10_mv1_v2_stride2_p01_val_0.0159/sample_002/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="10_mv1_v2_stride2_p01_val_0.0159/sample_002/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="10_mv1_v2_stride2_p01_val_0.0159/sample_002/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="10_mv1_v2_stride2_p01_val_0.0159_sample_003" class="sample"><summary><b>sample_003</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="10_mv1_v2_stride2_p01_val_0.0159/sample_003/panel.png" target="_blank"><img loading="lazy" src="10_mv1_v2_stride2_p01_val_0.0159/sample_003/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 10_mv1_v2_stride2_p01_val_0.0159/sample_003"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="10_mv1_v2_stride2_p01_val_0.0159/sample_003/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="10_mv1_v2_stride2_p01_val_0.0159/sample_003/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="10_mv1_v2_stride2_p01_val_0.0159/sample_003/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="10_mv1_v2_stride2_p01_val_0.0159/sample_003/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="10_mv1_v2_stride2_p01_val_0.0159/sample_003/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="10_mv1_v2_stride2_p01_val_0.0159_sample_004" class="sample"><summary><b>sample_004</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="10_mv1_v2_stride2_p01_val_0.0159/sample_004/panel.png" target="_blank"><img loading="lazy" src="10_mv1_v2_stride2_p01_val_0.0159/sample_004/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 10_mv1_v2_stride2_p01_val_0.0159/sample_004"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="10_mv1_v2_stride2_p01_val_0.0159/sample_004/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="10_mv1_v2_stride2_p01_val_0.0159/sample_004/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="10_mv1_v2_stride2_p01_val_0.0159/sample_004/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="10_mv1_v2_stride2_p01_val_0.0159/sample_004/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="10_mv1_v2_stride2_p01_val_0.0159/sample_004/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="10_mv1_v2_stride2_p01_val_0.0159_sample_005" class="sample"><summary><b>sample_005</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="10_mv1_v2_stride2_p01_val_0.0159/sample_005/panel.png" target="_blank"><img loading="lazy" src="10_mv1_v2_stride2_p01_val_0.0159/sample_005/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 10_mv1_v2_stride2_p01_val_0.0159/sample_005"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="10_mv1_v2_stride2_p01_val_0.0159/sample_005/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="10_mv1_v2_stride2_p01_val_0.0159/sample_005/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="10_mv1_v2_stride2_p01_val_0.0159/sample_005/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="10_mv1_v2_stride2_p01_val_0.0159/sample_005/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="10_mv1_v2_stride2_p01_val_0.0159/sample_005/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="10_mv1_v2_stride2_p01_val_0.0159_sample_006" class="sample"><summary><b>sample_006</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="10_mv1_v2_stride2_p01_val_0.0159/sample_006/panel.png" target="_blank"><img loading="lazy" src="10_mv1_v2_stride2_p01_val_0.0159/sample_006/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 10_mv1_v2_stride2_p01_val_0.0159/sample_006"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="10_mv1_v2_stride2_p01_val_0.0159/sample_006/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="10_mv1_v2_stride2_p01_val_0.0159/sample_006/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="10_mv1_v2_stride2_p01_val_0.0159/sample_006/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="10_mv1_v2_stride2_p01_val_0.0159/sample_006/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="10_mv1_v2_stride2_p01_val_0.0159/sample_006/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="10_mv1_v2_stride2_p01_val_0.0159_sample_007" class="sample"><summary><b>sample_007</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="10_mv1_v2_stride2_p01_val_0.0159/sample_007/panel.png" target="_blank"><img loading="lazy" src="10_mv1_v2_stride2_p01_val_0.0159/sample_007/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 10_mv1_v2_stride2_p01_val_0.0159/sample_007"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="10_mv1_v2_stride2_p01_val_0.0159/sample_007/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="10_mv1_v2_stride2_p01_val_0.0159/sample_007/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="10_mv1_v2_stride2_p01_val_0.0159/sample_007/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="10_mv1_v2_stride2_p01_val_0.0159/sample_007/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="10_mv1_v2_stride2_p01_val_0.0159/sample_007/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| </div> | |
| <h3 id="group_10r_mv1_v2_stride2_p01rand_val_0.0140">p01 + random no-contact ref</h3> | |
| <p class="run-desc">p01 recipe with a randomly sampled no-contact reference frame per window instead of the fixed episode_p01 frame. | |
| Showing 8 of 10 samples.</p> | |
| <div class="samples"> | |
| <details id="10r_mv1_v2_stride2_p01rand_val_0.0140_sample_000" class="sample"><summary><b>sample_000</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="10r_mv1_v2_stride2_p01rand_val_0.0140/sample_000/panel.png" target="_blank"><img loading="lazy" src="10r_mv1_v2_stride2_p01rand_val_0.0140/sample_000/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 10r_mv1_v2_stride2_p01rand_val_0.0140/sample_000"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="10r_mv1_v2_stride2_p01rand_val_0.0140/sample_000/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="10r_mv1_v2_stride2_p01rand_val_0.0140/sample_000/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="10r_mv1_v2_stride2_p01rand_val_0.0140/sample_000/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="10r_mv1_v2_stride2_p01rand_val_0.0140/sample_000/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="10r_mv1_v2_stride2_p01rand_val_0.0140/sample_000/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="10r_mv1_v2_stride2_p01rand_val_0.0140_sample_001" class="sample"><summary><b>sample_001</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="10r_mv1_v2_stride2_p01rand_val_0.0140/sample_001/panel.png" target="_blank"><img loading="lazy" src="10r_mv1_v2_stride2_p01rand_val_0.0140/sample_001/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 10r_mv1_v2_stride2_p01rand_val_0.0140/sample_001"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="10r_mv1_v2_stride2_p01rand_val_0.0140/sample_001/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="10r_mv1_v2_stride2_p01rand_val_0.0140/sample_001/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="10r_mv1_v2_stride2_p01rand_val_0.0140/sample_001/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="10r_mv1_v2_stride2_p01rand_val_0.0140/sample_001/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="10r_mv1_v2_stride2_p01rand_val_0.0140/sample_001/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="10r_mv1_v2_stride2_p01rand_val_0.0140_sample_002" class="sample"><summary><b>sample_002</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="10r_mv1_v2_stride2_p01rand_val_0.0140/sample_002/panel.png" target="_blank"><img loading="lazy" src="10r_mv1_v2_stride2_p01rand_val_0.0140/sample_002/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 10r_mv1_v2_stride2_p01rand_val_0.0140/sample_002"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="10r_mv1_v2_stride2_p01rand_val_0.0140/sample_002/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="10r_mv1_v2_stride2_p01rand_val_0.0140/sample_002/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="10r_mv1_v2_stride2_p01rand_val_0.0140/sample_002/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="10r_mv1_v2_stride2_p01rand_val_0.0140/sample_002/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="10r_mv1_v2_stride2_p01rand_val_0.0140/sample_002/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="10r_mv1_v2_stride2_p01rand_val_0.0140_sample_003" class="sample"><summary><b>sample_003</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="10r_mv1_v2_stride2_p01rand_val_0.0140/sample_003/panel.png" target="_blank"><img loading="lazy" src="10r_mv1_v2_stride2_p01rand_val_0.0140/sample_003/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 10r_mv1_v2_stride2_p01rand_val_0.0140/sample_003"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="10r_mv1_v2_stride2_p01rand_val_0.0140/sample_003/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="10r_mv1_v2_stride2_p01rand_val_0.0140/sample_003/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="10r_mv1_v2_stride2_p01rand_val_0.0140/sample_003/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="10r_mv1_v2_stride2_p01rand_val_0.0140/sample_003/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="10r_mv1_v2_stride2_p01rand_val_0.0140/sample_003/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="10r_mv1_v2_stride2_p01rand_val_0.0140_sample_004" class="sample"><summary><b>sample_004</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="10r_mv1_v2_stride2_p01rand_val_0.0140/sample_004/panel.png" target="_blank"><img loading="lazy" src="10r_mv1_v2_stride2_p01rand_val_0.0140/sample_004/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 10r_mv1_v2_stride2_p01rand_val_0.0140/sample_004"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="10r_mv1_v2_stride2_p01rand_val_0.0140/sample_004/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="10r_mv1_v2_stride2_p01rand_val_0.0140/sample_004/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="10r_mv1_v2_stride2_p01rand_val_0.0140/sample_004/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="10r_mv1_v2_stride2_p01rand_val_0.0140/sample_004/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="10r_mv1_v2_stride2_p01rand_val_0.0140/sample_004/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="10r_mv1_v2_stride2_p01rand_val_0.0140_sample_005" class="sample"><summary><b>sample_005</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="10r_mv1_v2_stride2_p01rand_val_0.0140/sample_005/panel.png" target="_blank"><img loading="lazy" src="10r_mv1_v2_stride2_p01rand_val_0.0140/sample_005/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 10r_mv1_v2_stride2_p01rand_val_0.0140/sample_005"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="10r_mv1_v2_stride2_p01rand_val_0.0140/sample_005/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="10r_mv1_v2_stride2_p01rand_val_0.0140/sample_005/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="10r_mv1_v2_stride2_p01rand_val_0.0140/sample_005/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="10r_mv1_v2_stride2_p01rand_val_0.0140/sample_005/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="10r_mv1_v2_stride2_p01rand_val_0.0140/sample_005/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="10r_mv1_v2_stride2_p01rand_val_0.0140_sample_006" class="sample"><summary><b>sample_006</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="10r_mv1_v2_stride2_p01rand_val_0.0140/sample_006/panel.png" target="_blank"><img loading="lazy" src="10r_mv1_v2_stride2_p01rand_val_0.0140/sample_006/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 10r_mv1_v2_stride2_p01rand_val_0.0140/sample_006"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="10r_mv1_v2_stride2_p01rand_val_0.0140/sample_006/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="10r_mv1_v2_stride2_p01rand_val_0.0140/sample_006/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="10r_mv1_v2_stride2_p01rand_val_0.0140/sample_006/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="10r_mv1_v2_stride2_p01rand_val_0.0140/sample_006/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="10r_mv1_v2_stride2_p01rand_val_0.0140/sample_006/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="10r_mv1_v2_stride2_p01rand_val_0.0140_sample_007" class="sample"><summary><b>sample_007</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="10r_mv1_v2_stride2_p01rand_val_0.0140/sample_007/panel.png" target="_blank"><img loading="lazy" src="10r_mv1_v2_stride2_p01rand_val_0.0140/sample_007/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 10r_mv1_v2_stride2_p01rand_val_0.0140/sample_007"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="10r_mv1_v2_stride2_p01rand_val_0.0140/sample_007/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="10r_mv1_v2_stride2_p01rand_val_0.0140/sample_007/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="10r_mv1_v2_stride2_p01rand_val_0.0140/sample_007/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="10r_mv1_v2_stride2_p01rand_val_0.0140/sample_007/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="10r_mv1_v2_stride2_p01rand_val_0.0140/sample_007/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| </div> | |
| <h2 id="cat_C._Architecture_—_MoT___hybrid" class="cat-header">C. Architecture — MoT / hybrid</h2> | |
| <h3 id="group_11_mot_vt_p01_val_0.0165">MoT V+T</h3> | |
| <p class="run-desc">Codex-designed MoT variant — per-modality QKV+MLP+adaLN with shared spatial attention pool. Functionally ≡ mmdit_mv. | |
| Showing 8 of 10 samples.</p> | |
| <div class="samples"> | |
| <details id="11_mot_vt_p01_val_0.0165_sample_000" class="sample"><summary><b>sample_000</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="11_mot_vt_p01_val_0.0165/sample_000/panel.png" target="_blank"><img loading="lazy" src="11_mot_vt_p01_val_0.0165/sample_000/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 11_mot_vt_p01_val_0.0165/sample_000"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="11_mot_vt_p01_val_0.0165/sample_000/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="11_mot_vt_p01_val_0.0165/sample_000/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="11_mot_vt_p01_val_0.0165/sample_000/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="11_mot_vt_p01_val_0.0165/sample_000/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="11_mot_vt_p01_val_0.0165/sample_000/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="11_mot_vt_p01_val_0.0165_sample_001" class="sample"><summary><b>sample_001</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="11_mot_vt_p01_val_0.0165/sample_001/panel.png" target="_blank"><img loading="lazy" src="11_mot_vt_p01_val_0.0165/sample_001/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 11_mot_vt_p01_val_0.0165/sample_001"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="11_mot_vt_p01_val_0.0165/sample_001/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="11_mot_vt_p01_val_0.0165/sample_001/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="11_mot_vt_p01_val_0.0165/sample_001/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="11_mot_vt_p01_val_0.0165/sample_001/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="11_mot_vt_p01_val_0.0165/sample_001/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="11_mot_vt_p01_val_0.0165_sample_002" class="sample"><summary><b>sample_002</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="11_mot_vt_p01_val_0.0165/sample_002/panel.png" target="_blank"><img loading="lazy" src="11_mot_vt_p01_val_0.0165/sample_002/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 11_mot_vt_p01_val_0.0165/sample_002"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="11_mot_vt_p01_val_0.0165/sample_002/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="11_mot_vt_p01_val_0.0165/sample_002/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="11_mot_vt_p01_val_0.0165/sample_002/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="11_mot_vt_p01_val_0.0165/sample_002/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="11_mot_vt_p01_val_0.0165/sample_002/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="11_mot_vt_p01_val_0.0165_sample_003" class="sample"><summary><b>sample_003</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="11_mot_vt_p01_val_0.0165/sample_003/panel.png" target="_blank"><img loading="lazy" src="11_mot_vt_p01_val_0.0165/sample_003/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 11_mot_vt_p01_val_0.0165/sample_003"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="11_mot_vt_p01_val_0.0165/sample_003/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="11_mot_vt_p01_val_0.0165/sample_003/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="11_mot_vt_p01_val_0.0165/sample_003/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="11_mot_vt_p01_val_0.0165/sample_003/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="11_mot_vt_p01_val_0.0165/sample_003/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="11_mot_vt_p01_val_0.0165_sample_004" class="sample"><summary><b>sample_004</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="11_mot_vt_p01_val_0.0165/sample_004/panel.png" target="_blank"><img loading="lazy" src="11_mot_vt_p01_val_0.0165/sample_004/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 11_mot_vt_p01_val_0.0165/sample_004"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="11_mot_vt_p01_val_0.0165/sample_004/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="11_mot_vt_p01_val_0.0165/sample_004/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="11_mot_vt_p01_val_0.0165/sample_004/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="11_mot_vt_p01_val_0.0165/sample_004/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="11_mot_vt_p01_val_0.0165/sample_004/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="11_mot_vt_p01_val_0.0165_sample_005" class="sample"><summary><b>sample_005</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="11_mot_vt_p01_val_0.0165/sample_005/panel.png" target="_blank"><img loading="lazy" src="11_mot_vt_p01_val_0.0165/sample_005/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 11_mot_vt_p01_val_0.0165/sample_005"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="11_mot_vt_p01_val_0.0165/sample_005/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="11_mot_vt_p01_val_0.0165/sample_005/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="11_mot_vt_p01_val_0.0165/sample_005/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="11_mot_vt_p01_val_0.0165/sample_005/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="11_mot_vt_p01_val_0.0165/sample_005/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="11_mot_vt_p01_val_0.0165_sample_006" class="sample"><summary><b>sample_006</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="11_mot_vt_p01_val_0.0165/sample_006/panel.png" target="_blank"><img loading="lazy" src="11_mot_vt_p01_val_0.0165/sample_006/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 11_mot_vt_p01_val_0.0165/sample_006"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="11_mot_vt_p01_val_0.0165/sample_006/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="11_mot_vt_p01_val_0.0165/sample_006/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="11_mot_vt_p01_val_0.0165/sample_006/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="11_mot_vt_p01_val_0.0165/sample_006/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="11_mot_vt_p01_val_0.0165/sample_006/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="11_mot_vt_p01_val_0.0165_sample_007" class="sample"><summary><b>sample_007</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="11_mot_vt_p01_val_0.0165/sample_007/panel.png" target="_blank"><img loading="lazy" src="11_mot_vt_p01_val_0.0165/sample_007/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 11_mot_vt_p01_val_0.0165/sample_007"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="11_mot_vt_p01_val_0.0165/sample_007/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="11_mot_vt_p01_val_0.0165/sample_007/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="11_mot_vt_p01_val_0.0165/sample_007/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="11_mot_vt_p01_val_0.0165/sample_007/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="11_mot_vt_p01_val_0.0165/sample_007/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| </div> | |
| <h3 id="group_11b_mot_vta_p01_val_0.0167">MoT V+T+Action</h3> | |
| <p class="run-desc">MoT-VT + 4 zero-init action-expert tokens joining the spatial pool. The gate never opened without a supervision signal — action tokens contributed nothing. | |
| Showing 8 of 10 samples.</p> | |
| <div class="samples"> | |
| <details id="11b_mot_vta_p01_val_0.0167_sample_000" class="sample"><summary><b>sample_000</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="11b_mot_vta_p01_val_0.0167/sample_000/panel.png" target="_blank"><img loading="lazy" src="11b_mot_vta_p01_val_0.0167/sample_000/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 11b_mot_vta_p01_val_0.0167/sample_000"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="11b_mot_vta_p01_val_0.0167/sample_000/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="11b_mot_vta_p01_val_0.0167/sample_000/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="11b_mot_vta_p01_val_0.0167/sample_000/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="11b_mot_vta_p01_val_0.0167/sample_000/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="11b_mot_vta_p01_val_0.0167/sample_000/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="11b_mot_vta_p01_val_0.0167_sample_001" class="sample"><summary><b>sample_001</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="11b_mot_vta_p01_val_0.0167/sample_001/panel.png" target="_blank"><img loading="lazy" src="11b_mot_vta_p01_val_0.0167/sample_001/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 11b_mot_vta_p01_val_0.0167/sample_001"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="11b_mot_vta_p01_val_0.0167/sample_001/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="11b_mot_vta_p01_val_0.0167/sample_001/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="11b_mot_vta_p01_val_0.0167/sample_001/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="11b_mot_vta_p01_val_0.0167/sample_001/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="11b_mot_vta_p01_val_0.0167/sample_001/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="11b_mot_vta_p01_val_0.0167_sample_002" class="sample"><summary><b>sample_002</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="11b_mot_vta_p01_val_0.0167/sample_002/panel.png" target="_blank"><img loading="lazy" src="11b_mot_vta_p01_val_0.0167/sample_002/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 11b_mot_vta_p01_val_0.0167/sample_002"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="11b_mot_vta_p01_val_0.0167/sample_002/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="11b_mot_vta_p01_val_0.0167/sample_002/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="11b_mot_vta_p01_val_0.0167/sample_002/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="11b_mot_vta_p01_val_0.0167/sample_002/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="11b_mot_vta_p01_val_0.0167/sample_002/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="11b_mot_vta_p01_val_0.0167_sample_003" class="sample"><summary><b>sample_003</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="11b_mot_vta_p01_val_0.0167/sample_003/panel.png" target="_blank"><img loading="lazy" src="11b_mot_vta_p01_val_0.0167/sample_003/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 11b_mot_vta_p01_val_0.0167/sample_003"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="11b_mot_vta_p01_val_0.0167/sample_003/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="11b_mot_vta_p01_val_0.0167/sample_003/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="11b_mot_vta_p01_val_0.0167/sample_003/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="11b_mot_vta_p01_val_0.0167/sample_003/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="11b_mot_vta_p01_val_0.0167/sample_003/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="11b_mot_vta_p01_val_0.0167_sample_004" class="sample"><summary><b>sample_004</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="11b_mot_vta_p01_val_0.0167/sample_004/panel.png" target="_blank"><img loading="lazy" src="11b_mot_vta_p01_val_0.0167/sample_004/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 11b_mot_vta_p01_val_0.0167/sample_004"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="11b_mot_vta_p01_val_0.0167/sample_004/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="11b_mot_vta_p01_val_0.0167/sample_004/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="11b_mot_vta_p01_val_0.0167/sample_004/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="11b_mot_vta_p01_val_0.0167/sample_004/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="11b_mot_vta_p01_val_0.0167/sample_004/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="11b_mot_vta_p01_val_0.0167_sample_005" class="sample"><summary><b>sample_005</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="11b_mot_vta_p01_val_0.0167/sample_005/panel.png" target="_blank"><img loading="lazy" src="11b_mot_vta_p01_val_0.0167/sample_005/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 11b_mot_vta_p01_val_0.0167/sample_005"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="11b_mot_vta_p01_val_0.0167/sample_005/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="11b_mot_vta_p01_val_0.0167/sample_005/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="11b_mot_vta_p01_val_0.0167/sample_005/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="11b_mot_vta_p01_val_0.0167/sample_005/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="11b_mot_vta_p01_val_0.0167/sample_005/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="11b_mot_vta_p01_val_0.0167_sample_006" class="sample"><summary><b>sample_006</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="11b_mot_vta_p01_val_0.0167/sample_006/panel.png" target="_blank"><img loading="lazy" src="11b_mot_vta_p01_val_0.0167/sample_006/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 11b_mot_vta_p01_val_0.0167/sample_006"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="11b_mot_vta_p01_val_0.0167/sample_006/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="11b_mot_vta_p01_val_0.0167/sample_006/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="11b_mot_vta_p01_val_0.0167/sample_006/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="11b_mot_vta_p01_val_0.0167/sample_006/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="11b_mot_vta_p01_val_0.0167/sample_006/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="11b_mot_vta_p01_val_0.0167_sample_007" class="sample"><summary><b>sample_007</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="11b_mot_vta_p01_val_0.0167/sample_007/panel.png" target="_blank"><img loading="lazy" src="11b_mot_vta_p01_val_0.0167/sample_007/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 11b_mot_vta_p01_val_0.0167/sample_007"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="11b_mot_vta_p01_val_0.0167/sample_007/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="11b_mot_vta_p01_val_0.0167/sample_007/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="11b_mot_vta_p01_val_0.0167/sample_007/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="11b_mot_vta_p01_val_0.0167/sample_007/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="11b_mot_vta_p01_val_0.0167/sample_007/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| </div> | |
| <h3 id="group_12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156">stride2 + p01 + hybrid + RoPE</h3> | |
| <p class="run-desc">Dual→single hybrid fusion (HunyuanVideo/FLUX style) with RoPE positional encoding. Warm-started from mv1_v2_stride2_p01. | |
| Showing 8 of 10 samples.</p> | |
| <div class="samples"> | |
| <details id="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156_sample_000" class="sample"><summary><b>sample_000</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_000/panel.png" target="_blank"><img loading="lazy" src="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_000/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_000"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_000/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_000/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_000/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_000/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_000/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156_sample_001" class="sample"><summary><b>sample_001</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_001/panel.png" target="_blank"><img loading="lazy" src="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_001/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_001"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_001/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_001/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_001/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_001/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_001/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156_sample_002" class="sample"><summary><b>sample_002</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_002/panel.png" target="_blank"><img loading="lazy" src="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_002/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_002"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_002/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_002/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_002/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_002/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_002/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156_sample_003" class="sample"><summary><b>sample_003</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_003/panel.png" target="_blank"><img loading="lazy" src="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_003/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_003"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_003/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_003/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_003/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_003/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_003/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156_sample_004" class="sample"><summary><b>sample_004</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_004/panel.png" target="_blank"><img loading="lazy" src="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_004/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_004"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_004/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_004/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_004/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_004/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_004/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156_sample_005" class="sample"><summary><b>sample_005</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_005/panel.png" target="_blank"><img loading="lazy" src="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_005/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_005"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_005/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_005/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_005/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_005/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_005/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156_sample_006" class="sample"><summary><b>sample_006</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_006/panel.png" target="_blank"><img loading="lazy" src="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_006/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_006"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_006/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_006/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_006/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_006/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_006/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156_sample_007" class="sample"><summary><b>sample_007</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_007/panel.png" target="_blank"><img loading="lazy" src="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_007/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_007"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_007/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_007/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_007/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_007/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="12h_mv1_v2_stride2_p01_hybrid_rope_val_0.0156/sample_007/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| </div> | |
| <h3 id="group_12hb_mv1_v2_stride2_p01_hybrid_val_0.0161">stride2 + p01 + hybrid (dual6+single7)</h3> | |
| <p class="run-desc">Dual→single hybrid fusion, 85M params (0.84x mmdit_mv). First 6 blocks per-modality, last 7 shared-QKV with per-modality adaLN. Warm-started from mv1_v2_stride2_p01. Ties the p01 baseline — architecture is not the bottleneck. | |
| Showing 8 of 10 samples.</p> | |
| <div class="samples"> | |
| <details id="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161_sample_000" class="sample"><summary><b>sample_000</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_000/panel.png" target="_blank"><img loading="lazy" src="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_000/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_000"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_000/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_000/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_000/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_000/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_000/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161_sample_001" class="sample"><summary><b>sample_001</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_001/panel.png" target="_blank"><img loading="lazy" src="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_001/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_001"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_001/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_001/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_001/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_001/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_001/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161_sample_002" class="sample"><summary><b>sample_002</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_002/panel.png" target="_blank"><img loading="lazy" src="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_002/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_002"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_002/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_002/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_002/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_002/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_002/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161_sample_003" class="sample"><summary><b>sample_003</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_003/panel.png" target="_blank"><img loading="lazy" src="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_003/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_003"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_003/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_003/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_003/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_003/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_003/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161_sample_004" class="sample"><summary><b>sample_004</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_004/panel.png" target="_blank"><img loading="lazy" src="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_004/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_004"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_004/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_004/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_004/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_004/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_004/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161_sample_005" class="sample"><summary><b>sample_005</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_005/panel.png" target="_blank"><img loading="lazy" src="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_005/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_005"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_005/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_005/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_005/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_005/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_005/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161_sample_006" class="sample"><summary><b>sample_006</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_006/panel.png" target="_blank"><img loading="lazy" src="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_006/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_006"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_006/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_006/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_006/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_006/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_006/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161_sample_007" class="sample"><summary><b>sample_007</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_007/panel.png" target="_blank"><img loading="lazy" src="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_007/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_007"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_007/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_007/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_007/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_007/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="12hb_mv1_v2_stride2_p01_hybrid_val_0.0161/sample_007/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| </div> | |
| <h3 id="group_12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159">stride2 + p01 + hybrid (deeper, depth14)</h3> | |
| <p class="run-desc">Deeper hybrid: depth=14 (+2 single-stream blocks), ~100M params (matches mmdit_mv budget). dual_depth=6 preserves warm-start. Ties the p01 baseline at 0.0159. | |
| Showing 8 of 10 samples.</p> | |
| <div class="samples"> | |
| <details id="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159_sample_000" class="sample"><summary><b>sample_000</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_000/panel.png" target="_blank"><img loading="lazy" src="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_000/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_000"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_000/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_000/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_000/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_000/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_000/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159_sample_001" class="sample"><summary><b>sample_001</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_001/panel.png" target="_blank"><img loading="lazy" src="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_001/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_001"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_001/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_001/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_001/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_001/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_001/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159_sample_002" class="sample"><summary><b>sample_002</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_002/panel.png" target="_blank"><img loading="lazy" src="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_002/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_002"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_002/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_002/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_002/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_002/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_002/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159_sample_003" class="sample"><summary><b>sample_003</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_003/panel.png" target="_blank"><img loading="lazy" src="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_003/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_003"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_003/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_003/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_003/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_003/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_003/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159_sample_004" class="sample"><summary><b>sample_004</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_004/panel.png" target="_blank"><img loading="lazy" src="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_004/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_004"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_004/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_004/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_004/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_004/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_004/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159_sample_005" class="sample"><summary><b>sample_005</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_005/panel.png" target="_blank"><img loading="lazy" src="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_005/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_005"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_005/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_005/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_005/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_005/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_005/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159_sample_006" class="sample"><summary><b>sample_006</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_006/panel.png" target="_blank"><img loading="lazy" src="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_006/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_006"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_006/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_006/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_006/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_006/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_006/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159_sample_007" class="sample"><summary><b>sample_007</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_007/panel.png" target="_blank"><img loading="lazy" src="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_007/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_007"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_007/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_007/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_007/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_007/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="12hd_mv1_v2_stride2_p01_hybrid_deep_val_0.0159/sample_007/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| </div> | |
| <h2 id="cat_D._Single-modality_controls" class="cat-header">D. Single-modality controls</h2> | |
| <h3 id="group_12v_mv1_v2_stride2_p01_vision_only_val_0.0093">vision-only control</h3> | |
| <p class="run-desc">p01 recipe with the tactile stream removed — visual-only lower bound (val_loss 0.0101). | |
| Showing 8 of 10 samples.</p> | |
| <div class="samples"> | |
| <details id="12v_mv1_v2_stride2_p01_vision_only_val_0.0093_sample_000" class="sample"><summary><b>sample_000</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_000/panel.png" target="_blank"><img loading="lazy" src="12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_000/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_000"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_000/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_000/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_000/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_000/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_000/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="12v_mv1_v2_stride2_p01_vision_only_val_0.0093_sample_001" class="sample"><summary><b>sample_001</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_001/panel.png" target="_blank"><img loading="lazy" src="12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_001/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_001"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_001/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_001/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_001/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_001/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_001/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="12v_mv1_v2_stride2_p01_vision_only_val_0.0093_sample_002" class="sample"><summary><b>sample_002</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_002/panel.png" target="_blank"><img loading="lazy" src="12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_002/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_002"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_002/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_002/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_002/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_002/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_002/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="12v_mv1_v2_stride2_p01_vision_only_val_0.0093_sample_003" class="sample"><summary><b>sample_003</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_003/panel.png" target="_blank"><img loading="lazy" src="12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_003/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_003"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_003/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_003/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_003/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_003/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_003/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="12v_mv1_v2_stride2_p01_vision_only_val_0.0093_sample_004" class="sample"><summary><b>sample_004</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_004/panel.png" target="_blank"><img loading="lazy" src="12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_004/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_004"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_004/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_004/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_004/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_004/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_004/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="12v_mv1_v2_stride2_p01_vision_only_val_0.0093_sample_005" class="sample"><summary><b>sample_005</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_005/panel.png" target="_blank"><img loading="lazy" src="12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_005/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_005"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_005/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_005/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_005/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_005/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_005/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="12v_mv1_v2_stride2_p01_vision_only_val_0.0093_sample_006" class="sample"><summary><b>sample_006</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_006/panel.png" target="_blank"><img loading="lazy" src="12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_006/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_006"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_006/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_006/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_006/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_006/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_006/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="12v_mv1_v2_stride2_p01_vision_only_val_0.0093_sample_007" class="sample"><summary><b>sample_007</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_007/panel.png" target="_blank"><img loading="lazy" src="12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_007/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_007"/></a></div><div class="grid"><figure><figcaption>view_left — GT | rollout</figcaption><video src="12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_007/view_left_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_middle (top-down) — GT | rollout</figcaption><video src="12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_007/view_middle_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>view_right — GT | rollout</figcaption><video src="12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_007/view_right_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_007/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="12v_mv1_v2_stride2_p01_vision_only_val_0.0093/sample_007/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| </div> | |
| <h3 id="group_13t_mv1_v2_stride2_p01_tactile_only_val_0.0062">tactile-only control</h3> | |
| <p class="run-desc">p01 recipe with the visual stream removed — tactile-only lower bound (val_loss 0.0062; tactile latents are easier to predict than visual). | |
| Showing 8 of 10 samples.</p> | |
| <div class="samples"> | |
| <details id="13t_mv1_v2_stride2_p01_tactile_only_val_0.0062_sample_000" class="sample"><summary><b>sample_000</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="13t_mv1_v2_stride2_p01_tactile_only_val_0.0062/sample_000/panel.png" target="_blank"><img loading="lazy" src="13t_mv1_v2_stride2_p01_tactile_only_val_0.0062/sample_000/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 13t_mv1_v2_stride2_p01_tactile_only_val_0.0062/sample_000"/></a></div><div class="grid"><figure><figcaption>view — GT | rollout</figcaption><video src="13t_mv1_v2_stride2_p01_tactile_only_val_0.0062/sample_000/view_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="13t_mv1_v2_stride2_p01_tactile_only_val_0.0062/sample_000/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="13t_mv1_v2_stride2_p01_tactile_only_val_0.0062/sample_000/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="13t_mv1_v2_stride2_p01_tactile_only_val_0.0062_sample_001" class="sample"><summary><b>sample_001</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="13t_mv1_v2_stride2_p01_tactile_only_val_0.0062/sample_001/panel.png" target="_blank"><img loading="lazy" src="13t_mv1_v2_stride2_p01_tactile_only_val_0.0062/sample_001/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 13t_mv1_v2_stride2_p01_tactile_only_val_0.0062/sample_001"/></a></div><div class="grid"><figure><figcaption>view — GT | rollout</figcaption><video src="13t_mv1_v2_stride2_p01_tactile_only_val_0.0062/sample_001/view_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="13t_mv1_v2_stride2_p01_tactile_only_val_0.0062/sample_001/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="13t_mv1_v2_stride2_p01_tactile_only_val_0.0062/sample_001/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="13t_mv1_v2_stride2_p01_tactile_only_val_0.0062_sample_002" class="sample"><summary><b>sample_002</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="13t_mv1_v2_stride2_p01_tactile_only_val_0.0062/sample_002/panel.png" target="_blank"><img loading="lazy" src="13t_mv1_v2_stride2_p01_tactile_only_val_0.0062/sample_002/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 13t_mv1_v2_stride2_p01_tactile_only_val_0.0062/sample_002"/></a></div><div class="grid"><figure><figcaption>view — GT | rollout</figcaption><video src="13t_mv1_v2_stride2_p01_tactile_only_val_0.0062/sample_002/view_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="13t_mv1_v2_stride2_p01_tactile_only_val_0.0062/sample_002/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="13t_mv1_v2_stride2_p01_tactile_only_val_0.0062/sample_002/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="13t_mv1_v2_stride2_p01_tactile_only_val_0.0062_sample_003" class="sample"><summary><b>sample_003</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="13t_mv1_v2_stride2_p01_tactile_only_val_0.0062/sample_003/panel.png" target="_blank"><img loading="lazy" src="13t_mv1_v2_stride2_p01_tactile_only_val_0.0062/sample_003/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 13t_mv1_v2_stride2_p01_tactile_only_val_0.0062/sample_003"/></a></div><div class="grid"><figure><figcaption>view — GT | rollout</figcaption><video src="13t_mv1_v2_stride2_p01_tactile_only_val_0.0062/sample_003/view_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="13t_mv1_v2_stride2_p01_tactile_only_val_0.0062/sample_003/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="13t_mv1_v2_stride2_p01_tactile_only_val_0.0062/sample_003/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="13t_mv1_v2_stride2_p01_tactile_only_val_0.0062_sample_004" class="sample"><summary><b>sample_004</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="13t_mv1_v2_stride2_p01_tactile_only_val_0.0062/sample_004/panel.png" target="_blank"><img loading="lazy" src="13t_mv1_v2_stride2_p01_tactile_only_val_0.0062/sample_004/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 13t_mv1_v2_stride2_p01_tactile_only_val_0.0062/sample_004"/></a></div><div class="grid"><figure><figcaption>view — GT | rollout</figcaption><video src="13t_mv1_v2_stride2_p01_tactile_only_val_0.0062/sample_004/view_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="13t_mv1_v2_stride2_p01_tactile_only_val_0.0062/sample_004/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="13t_mv1_v2_stride2_p01_tactile_only_val_0.0062/sample_004/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="13t_mv1_v2_stride2_p01_tactile_only_val_0.0062_sample_005" class="sample"><summary><b>sample_005</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="13t_mv1_v2_stride2_p01_tactile_only_val_0.0062/sample_005/panel.png" target="_blank"><img loading="lazy" src="13t_mv1_v2_stride2_p01_tactile_only_val_0.0062/sample_005/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 13t_mv1_v2_stride2_p01_tactile_only_val_0.0062/sample_005"/></a></div><div class="grid"><figure><figcaption>view — GT | rollout</figcaption><video src="13t_mv1_v2_stride2_p01_tactile_only_val_0.0062/sample_005/view_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="13t_mv1_v2_stride2_p01_tactile_only_val_0.0062/sample_005/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="13t_mv1_v2_stride2_p01_tactile_only_val_0.0062/sample_005/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="13t_mv1_v2_stride2_p01_tactile_only_val_0.0062_sample_006" class="sample"><summary><b>sample_006</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="13t_mv1_v2_stride2_p01_tactile_only_val_0.0062/sample_006/panel.png" target="_blank"><img loading="lazy" src="13t_mv1_v2_stride2_p01_tactile_only_val_0.0062/sample_006/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 13t_mv1_v2_stride2_p01_tactile_only_val_0.0062/sample_006"/></a></div><div class="grid"><figure><figcaption>view — GT | rollout</figcaption><video src="13t_mv1_v2_stride2_p01_tactile_only_val_0.0062/sample_006/view_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="13t_mv1_v2_stride2_p01_tactile_only_val_0.0062/sample_006/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="13t_mv1_v2_stride2_p01_tactile_only_val_0.0062/sample_006/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="13t_mv1_v2_stride2_p01_tactile_only_val_0.0062_sample_007" class="sample"><summary><b>sample_007</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="13t_mv1_v2_stride2_p01_tactile_only_val_0.0062/sample_007/panel.png" target="_blank"><img loading="lazy" src="13t_mv1_v2_stride2_p01_tactile_only_val_0.0062/sample_007/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 13t_mv1_v2_stride2_p01_tactile_only_val_0.0062/sample_007"/></a></div><div class="grid"><figure><figcaption>view — GT | rollout</figcaption><video src="13t_mv1_v2_stride2_p01_tactile_only_val_0.0062/sample_007/view_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="13t_mv1_v2_stride2_p01_tactile_only_val_0.0062/sample_007/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="13t_mv1_v2_stride2_p01_tactile_only_val_0.0062/sample_007/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| </div> | |
| <h2 id="cat_E._Cosmos_tokenizer" class="cat-header">E. Cosmos tokenizer</h2> | |
| <h3 id="group_20c_mv1_v2_stride2_p01_cosmos_v3_causal_val_0.1647">Cosmos v3 + causal backbone</h3> | |
| <p class="run-desc">Cosmos CV4x8x8 latents (v3 delta-space per-channel whitening) with a CAUSAL diffusion backbone (action_mode=per_frame_causal) to match the causal tokenizer + AR rollout; anchor_loss_weight=1.0. val_loss is on the Cosmos/whitened scale — NOT comparable to the VideoVAE+ runs above; judge by decoded rollout quality only. | |
| Showing 8 of 16 samples.</p> | |
| <div class="samples"> | |
| <details id="20c_mv1_v2_stride2_p01_cosmos_v3_causal_val_0.1647_sample_000" class="sample"><summary><b>sample_000</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="20c_mv1_v2_stride2_p01_cosmos_v3_causal_val_0.1647/sample_000/panel.png" target="_blank"><img loading="lazy" src="20c_mv1_v2_stride2_p01_cosmos_v3_causal_val_0.1647/sample_000/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 20c_mv1_v2_stride2_p01_cosmos_v3_causal_val_0.1647/sample_000"/></a></div><div class="grid"><figure><figcaption>view — GT | rollout</figcaption><video src="20c_mv1_v2_stride2_p01_cosmos_v3_causal_val_0.1647/sample_000/view_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="20c_mv1_v2_stride2_p01_cosmos_v3_causal_val_0.1647/sample_000/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="20c_mv1_v2_stride2_p01_cosmos_v3_causal_val_0.1647/sample_000/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="20c_mv1_v2_stride2_p01_cosmos_v3_causal_val_0.1647_sample_001" class="sample"><summary><b>sample_001</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="20c_mv1_v2_stride2_p01_cosmos_v3_causal_val_0.1647/sample_001/panel.png" target="_blank"><img loading="lazy" src="20c_mv1_v2_stride2_p01_cosmos_v3_causal_val_0.1647/sample_001/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 20c_mv1_v2_stride2_p01_cosmos_v3_causal_val_0.1647/sample_001"/></a></div><div class="grid"><figure><figcaption>view — GT | rollout</figcaption><video src="20c_mv1_v2_stride2_p01_cosmos_v3_causal_val_0.1647/sample_001/view_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="20c_mv1_v2_stride2_p01_cosmos_v3_causal_val_0.1647/sample_001/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="20c_mv1_v2_stride2_p01_cosmos_v3_causal_val_0.1647/sample_001/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="20c_mv1_v2_stride2_p01_cosmos_v3_causal_val_0.1647_sample_002" class="sample"><summary><b>sample_002</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="20c_mv1_v2_stride2_p01_cosmos_v3_causal_val_0.1647/sample_002/panel.png" target="_blank"><img loading="lazy" src="20c_mv1_v2_stride2_p01_cosmos_v3_causal_val_0.1647/sample_002/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 20c_mv1_v2_stride2_p01_cosmos_v3_causal_val_0.1647/sample_002"/></a></div><div class="grid"><figure><figcaption>view — GT | rollout</figcaption><video src="20c_mv1_v2_stride2_p01_cosmos_v3_causal_val_0.1647/sample_002/view_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="20c_mv1_v2_stride2_p01_cosmos_v3_causal_val_0.1647/sample_002/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="20c_mv1_v2_stride2_p01_cosmos_v3_causal_val_0.1647/sample_002/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="20c_mv1_v2_stride2_p01_cosmos_v3_causal_val_0.1647_sample_003" class="sample"><summary><b>sample_003</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="20c_mv1_v2_stride2_p01_cosmos_v3_causal_val_0.1647/sample_003/panel.png" target="_blank"><img loading="lazy" src="20c_mv1_v2_stride2_p01_cosmos_v3_causal_val_0.1647/sample_003/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 20c_mv1_v2_stride2_p01_cosmos_v3_causal_val_0.1647/sample_003"/></a></div><div class="grid"><figure><figcaption>view — GT | rollout</figcaption><video src="20c_mv1_v2_stride2_p01_cosmos_v3_causal_val_0.1647/sample_003/view_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="20c_mv1_v2_stride2_p01_cosmos_v3_causal_val_0.1647/sample_003/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="20c_mv1_v2_stride2_p01_cosmos_v3_causal_val_0.1647/sample_003/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="20c_mv1_v2_stride2_p01_cosmos_v3_causal_val_0.1647_sample_004" class="sample"><summary><b>sample_004</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="20c_mv1_v2_stride2_p01_cosmos_v3_causal_val_0.1647/sample_004/panel.png" target="_blank"><img loading="lazy" src="20c_mv1_v2_stride2_p01_cosmos_v3_causal_val_0.1647/sample_004/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 20c_mv1_v2_stride2_p01_cosmos_v3_causal_val_0.1647/sample_004"/></a></div><div class="grid"><figure><figcaption>view — GT | rollout</figcaption><video src="20c_mv1_v2_stride2_p01_cosmos_v3_causal_val_0.1647/sample_004/view_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="20c_mv1_v2_stride2_p01_cosmos_v3_causal_val_0.1647/sample_004/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="20c_mv1_v2_stride2_p01_cosmos_v3_causal_val_0.1647/sample_004/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="20c_mv1_v2_stride2_p01_cosmos_v3_causal_val_0.1647_sample_005" class="sample"><summary><b>sample_005</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="20c_mv1_v2_stride2_p01_cosmos_v3_causal_val_0.1647/sample_005/panel.png" target="_blank"><img loading="lazy" src="20c_mv1_v2_stride2_p01_cosmos_v3_causal_val_0.1647/sample_005/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 20c_mv1_v2_stride2_p01_cosmos_v3_causal_val_0.1647/sample_005"/></a></div><div class="grid"><figure><figcaption>view — GT | rollout</figcaption><video src="20c_mv1_v2_stride2_p01_cosmos_v3_causal_val_0.1647/sample_005/view_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="20c_mv1_v2_stride2_p01_cosmos_v3_causal_val_0.1647/sample_005/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="20c_mv1_v2_stride2_p01_cosmos_v3_causal_val_0.1647/sample_005/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="20c_mv1_v2_stride2_p01_cosmos_v3_causal_val_0.1647_sample_006" class="sample"><summary><b>sample_006</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="20c_mv1_v2_stride2_p01_cosmos_v3_causal_val_0.1647/sample_006/panel.png" target="_blank"><img loading="lazy" src="20c_mv1_v2_stride2_p01_cosmos_v3_causal_val_0.1647/sample_006/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 20c_mv1_v2_stride2_p01_cosmos_v3_causal_val_0.1647/sample_006"/></a></div><div class="grid"><figure><figcaption>view — GT | rollout</figcaption><video src="20c_mv1_v2_stride2_p01_cosmos_v3_causal_val_0.1647/sample_006/view_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="20c_mv1_v2_stride2_p01_cosmos_v3_causal_val_0.1647/sample_006/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="20c_mv1_v2_stride2_p01_cosmos_v3_causal_val_0.1647/sample_006/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="20c_mv1_v2_stride2_p01_cosmos_v3_causal_val_0.1647_sample_007" class="sample"><summary><b>sample_007</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="20c_mv1_v2_stride2_p01_cosmos_v3_causal_val_0.1647/sample_007/panel.png" target="_blank"><img loading="lazy" src="20c_mv1_v2_stride2_p01_cosmos_v3_causal_val_0.1647/sample_007/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 20c_mv1_v2_stride2_p01_cosmos_v3_causal_val_0.1647/sample_007"/></a></div><div class="grid"><figure><figcaption>view — GT | rollout</figcaption><video src="20c_mv1_v2_stride2_p01_cosmos_v3_causal_val_0.1647/sample_007/view_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="20c_mv1_v2_stride2_p01_cosmos_v3_causal_val_0.1647/sample_007/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="20c_mv1_v2_stride2_p01_cosmos_v3_causal_val_0.1647/sample_007/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| </div> | |
| <h3 id="group_20d_mv1_v2_stride2_p01_cosmos_v3_contact_val_0.1301">Cosmos v3 + contact-aware tactile loss</h3> | |
| <p class="run-desc">v3 + additive x0-space contact-aware tactile aux losses (contact-patch reconstruction 0.25 + false-contact suppression 0.10, alpha_bar-weighted; docs/CONTACT_AWARE_TACTILE_LOSS.md). Thresholds calibrated for Cosmos delta units (0.35/0.076/0.11 = P95/P70/P80 pixel energy). Single-variable vs v3 (full_window backbone, anchor 0.5). Judge by tactile contact MSE + false-contact rate + view LPIPS, not val_loss. | |
| Showing 8 of 16 samples.</p> | |
| <div class="samples"> | |
| <details id="20d_mv1_v2_stride2_p01_cosmos_v3_contact_val_0.1301_sample_000" class="sample"><summary><b>sample_000</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="20d_mv1_v2_stride2_p01_cosmos_v3_contact_val_0.1301/sample_000/panel.png" target="_blank"><img loading="lazy" src="20d_mv1_v2_stride2_p01_cosmos_v3_contact_val_0.1301/sample_000/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 20d_mv1_v2_stride2_p01_cosmos_v3_contact_val_0.1301/sample_000"/></a></div><div class="grid"><figure><figcaption>view — GT | rollout</figcaption><video src="20d_mv1_v2_stride2_p01_cosmos_v3_contact_val_0.1301/sample_000/view_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="20d_mv1_v2_stride2_p01_cosmos_v3_contact_val_0.1301/sample_000/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="20d_mv1_v2_stride2_p01_cosmos_v3_contact_val_0.1301/sample_000/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="20d_mv1_v2_stride2_p01_cosmos_v3_contact_val_0.1301_sample_001" class="sample"><summary><b>sample_001</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="20d_mv1_v2_stride2_p01_cosmos_v3_contact_val_0.1301/sample_001/panel.png" target="_blank"><img loading="lazy" src="20d_mv1_v2_stride2_p01_cosmos_v3_contact_val_0.1301/sample_001/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 20d_mv1_v2_stride2_p01_cosmos_v3_contact_val_0.1301/sample_001"/></a></div><div class="grid"><figure><figcaption>view — GT | rollout</figcaption><video src="20d_mv1_v2_stride2_p01_cosmos_v3_contact_val_0.1301/sample_001/view_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="20d_mv1_v2_stride2_p01_cosmos_v3_contact_val_0.1301/sample_001/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="20d_mv1_v2_stride2_p01_cosmos_v3_contact_val_0.1301/sample_001/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="20d_mv1_v2_stride2_p01_cosmos_v3_contact_val_0.1301_sample_002" class="sample"><summary><b>sample_002</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="20d_mv1_v2_stride2_p01_cosmos_v3_contact_val_0.1301/sample_002/panel.png" target="_blank"><img loading="lazy" src="20d_mv1_v2_stride2_p01_cosmos_v3_contact_val_0.1301/sample_002/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 20d_mv1_v2_stride2_p01_cosmos_v3_contact_val_0.1301/sample_002"/></a></div><div class="grid"><figure><figcaption>view — GT | rollout</figcaption><video src="20d_mv1_v2_stride2_p01_cosmos_v3_contact_val_0.1301/sample_002/view_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="20d_mv1_v2_stride2_p01_cosmos_v3_contact_val_0.1301/sample_002/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="20d_mv1_v2_stride2_p01_cosmos_v3_contact_val_0.1301/sample_002/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="20d_mv1_v2_stride2_p01_cosmos_v3_contact_val_0.1301_sample_003" class="sample"><summary><b>sample_003</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="20d_mv1_v2_stride2_p01_cosmos_v3_contact_val_0.1301/sample_003/panel.png" target="_blank"><img loading="lazy" src="20d_mv1_v2_stride2_p01_cosmos_v3_contact_val_0.1301/sample_003/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 20d_mv1_v2_stride2_p01_cosmos_v3_contact_val_0.1301/sample_003"/></a></div><div class="grid"><figure><figcaption>view — GT | rollout</figcaption><video src="20d_mv1_v2_stride2_p01_cosmos_v3_contact_val_0.1301/sample_003/view_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="20d_mv1_v2_stride2_p01_cosmos_v3_contact_val_0.1301/sample_003/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="20d_mv1_v2_stride2_p01_cosmos_v3_contact_val_0.1301/sample_003/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="20d_mv1_v2_stride2_p01_cosmos_v3_contact_val_0.1301_sample_004" class="sample"><summary><b>sample_004</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="20d_mv1_v2_stride2_p01_cosmos_v3_contact_val_0.1301/sample_004/panel.png" target="_blank"><img loading="lazy" src="20d_mv1_v2_stride2_p01_cosmos_v3_contact_val_0.1301/sample_004/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 20d_mv1_v2_stride2_p01_cosmos_v3_contact_val_0.1301/sample_004"/></a></div><div class="grid"><figure><figcaption>view — GT | rollout</figcaption><video src="20d_mv1_v2_stride2_p01_cosmos_v3_contact_val_0.1301/sample_004/view_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="20d_mv1_v2_stride2_p01_cosmos_v3_contact_val_0.1301/sample_004/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="20d_mv1_v2_stride2_p01_cosmos_v3_contact_val_0.1301/sample_004/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="20d_mv1_v2_stride2_p01_cosmos_v3_contact_val_0.1301_sample_005" class="sample"><summary><b>sample_005</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="20d_mv1_v2_stride2_p01_cosmos_v3_contact_val_0.1301/sample_005/panel.png" target="_blank"><img loading="lazy" src="20d_mv1_v2_stride2_p01_cosmos_v3_contact_val_0.1301/sample_005/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 20d_mv1_v2_stride2_p01_cosmos_v3_contact_val_0.1301/sample_005"/></a></div><div class="grid"><figure><figcaption>view — GT | rollout</figcaption><video src="20d_mv1_v2_stride2_p01_cosmos_v3_contact_val_0.1301/sample_005/view_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="20d_mv1_v2_stride2_p01_cosmos_v3_contact_val_0.1301/sample_005/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="20d_mv1_v2_stride2_p01_cosmos_v3_contact_val_0.1301/sample_005/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="20d_mv1_v2_stride2_p01_cosmos_v3_contact_val_0.1301_sample_006" class="sample"><summary><b>sample_006</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="20d_mv1_v2_stride2_p01_cosmos_v3_contact_val_0.1301/sample_006/panel.png" target="_blank"><img loading="lazy" src="20d_mv1_v2_stride2_p01_cosmos_v3_contact_val_0.1301/sample_006/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 20d_mv1_v2_stride2_p01_cosmos_v3_contact_val_0.1301/sample_006"/></a></div><div class="grid"><figure><figcaption>view — GT | rollout</figcaption><video src="20d_mv1_v2_stride2_p01_cosmos_v3_contact_val_0.1301/sample_006/view_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="20d_mv1_v2_stride2_p01_cosmos_v3_contact_val_0.1301/sample_006/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="20d_mv1_v2_stride2_p01_cosmos_v3_contact_val_0.1301/sample_006/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| <details id="20d_mv1_v2_stride2_p01_cosmos_v3_contact_val_0.1301_sample_007" class="sample"><summary><b>sample_007</b> <span style="color:#888;font-size:12px">(click to expand videos)</span></summary><div class="panel"><a href="20d_mv1_v2_stride2_p01_cosmos_v3_contact_val_0.1301/sample_007/panel.png" target="_blank"><img loading="lazy" src="20d_mv1_v2_stride2_p01_cosmos_v3_contact_val_0.1301/sample_007/panel.png" style="max-width:100%;max-height:240px;border:1px solid #ccc" alt="panel for 20d_mv1_v2_stride2_p01_cosmos_v3_contact_val_0.1301/sample_007"/></a></div><div class="grid"><figure><figcaption>view — GT | rollout</figcaption><video src="20d_mv1_v2_stride2_p01_cosmos_v3_contact_val_0.1301/sample_007/view_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_left — GT | rollout</figcaption><video src="20d_mv1_v2_stride2_p01_cosmos_v3_contact_val_0.1301/sample_007/tl_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure><figure><figcaption>tactile_right — GT | rollout</figcaption><video src="20d_mv1_v2_stride2_p01_cosmos_v3_contact_val_0.1301/sample_007/tr_gt_vs_rollout.mp4" loop muted playsinline controls preload="none" width="384"></video></figure></div></details> | |
| </div> | |
| </body></html> |