Add files using upload-large-folder tool

results/omni_finetune/xperience10m_qwen3_omni_v6_cross_modal_retrieval_probe_a100_20260618T000000Z/RUN_REPORT.md

@@ -0,0 +1,8 @@
+# Qwen3-Omni v6 Retrieval Task Probes
+
+- Run ID: `xperience10m_qwen3_omni_v6_cross_modal_retrieval_probe_a100_20260618T000000Z`
+- Shards: `2`
+
+| Task | Metric | Score | Samples |
+| --- | --- | ---: | ---: |
+| Cross-Modal Retrieval | cross_modal_retrieval_mrr | 0.508019 | 4032 |

results/omni_finetune/xperience10m_qwen3_omni_v6_cross_modal_retrieval_probe_a100_20260618T000000Z/collection_validation.json

@@ -0,0 +1,16 @@
+{
+  "records": [
+    {
+      "metric_key": "cross_modal_retrieval_mrr",
+      "num_samples": 4032,
+      "primary_score": 0.5080191798941732,
+      "source": "results/omni_finetune/xperience10m_qwen3_omni_v6_cross_modal_retrieval_probe_a100_20260618T000000Z/cross_modal_retrieval/metrics.json",
+      "task_id": "cross_modal_retrieval"
+    }
+  ],
+  "run_id": "xperience10m_qwen3_omni_v6_cross_modal_retrieval_probe_a100_20260618T000000Z",
+  "status": "pass",
+  "summary": "results/omni_finetune/xperience10m_qwen3_omni_v6_cross_modal_retrieval_probe_a100_20260618T000000Z/summary.json",
+  "title": "Qwen3 Retrieval Task Probe Collection Validation",
+  "validated_task_count": 1
+}

results/omni_finetune/xperience10m_qwen3_omni_v6_cross_modal_retrieval_probe_a100_20260618T000000Z/cross_modal_retrieval/metrics.json

@@ -0,0 +1,25 @@
+{
+  "adapter_dir": "/mnt/kgc/chaoyue/ropedia-h20-side/ropedia-episode-task-suite/checkpoints/xperience10m_qwen3_omni_128ep_multiscale_cap96_v6_rank64_lr5e5_full8gpu_lora/adapter_lora",
+  "candidate_count": 4,
+  "caption_grounding_mrr": 0.5080191798941732,
+  "cross_modal_retrieval_mrr": 0.5080191798941732,
+  "dataset_jsonl": "/mnt/kgc/chaoyue/ropedia-h20-side/ropedia-episode-task-suite/results/omni_finetune/xperience10m_qwen3_omni_128ep_multiscale_cap96_v5_full8gpu_lora_dataset/dataset_a100_eval.jsonl",
+  "eval_split": "test",
+  "metric_key": "cross_modal_retrieval_mrr",
+  "model_id": "/mnt/kgc/chaoyue/ropedia-h20-side/modelscope_models/Qwen__Qwen3-Omni-30B-A3B-Instruct",
+  "mrr": 0.5080191798941732,
+  "num_samples": 4032,
+  "primary_metric": "cross_modal_retrieval_mrr",
+  "primary_score": 0.5080191798941732,
+  "run_id": "xperience10m_qwen3_omni_v6_cross_modal_retrieval_probe_a100_20260618T000000Z",
+  "sample_offset": 0,
+  "sample_stride": 1,
+  "scope": "held_out_test_qwen3_retrieval_task_probe",
+  "score_policy": "GPU-backed Qwen3-Omni v6 sensor-to-video retrieval probe. The query is a compact summary of held-out motion-capture, body-contact, camera-pose, and IMU feature blocks; candidates are shuffled staged mosaic video windows, and the score is MRR of the synchronized true window. No action/subtask/object labels are included in the query.",
+  "status": "pass",
+  "task_id": "cross_modal_retrieval",
+  "task_label": "Cross-Modal Retrieval",
+  "task_number": 9,
+  "title": "Qwen3-Omni v6 Cross-Modal Retrieval",
+  "top1_accuracy": 0.24032738095238096
+}

results/omni_finetune/xperience10m_qwen3_omni_v6_cross_modal_retrieval_probe_a100_20260618T000000Z/launch_env.txt

@@ -0,0 +1,11 @@
+run_id=xperience10m_qwen3_omni_v6_cross_modal_retrieval_probe_a100_20260618T000000Z
+dataset_jsonl=/mnt/kgc/chaoyue/ropedia-h20-side/ropedia-episode-task-suite/results/omni_finetune/xperience10m_qwen3_omni_128ep_multiscale_cap96_v5_full8gpu_lora_dataset/dataset_a100_eval.jsonl
+model_dir=/mnt/kgc/chaoyue/ropedia-h20-side/modelscope_models/Qwen__Qwen3-Omni-30B-A3B-Instruct
+adapter_dir=/mnt/kgc/chaoyue/ropedia-h20-side/ropedia-episode-task-suite/checkpoints/xperience10m_qwen3_omni_128ep_multiscale_cap96_v6_rank64_lr5e5_full8gpu_lora/adapter_lora
+tasks=cross_modal_retrieval
+candidate_count=4
+cuda_device_groups=0,1 2,3
+shards=2
+started_at=2026-06-18T14:33:01+08:00
+finished_at=2026-06-18T18:01:20+08:00
+exit_code=0

results/omni_finetune/xperience10m_qwen3_omni_v6_cross_modal_retrieval_probe_a100_20260618T000000Z/summary.json

@@ -0,0 +1,19 @@
+{
+  "run_id": "xperience10m_qwen3_omni_v6_cross_modal_retrieval_probe_a100_20260618T000000Z",
+  "shard_dirs": [
+    "results/omni_finetune/xperience10m_qwen3_omni_v6_cross_modal_retrieval_probe_a100_20260618T000000Z_shard0",
+    "results/omni_finetune/xperience10m_qwen3_omni_v6_cross_modal_retrieval_probe_a100_20260618T000000Z_shard1"
+  ],
+  "status": "pass",
+  "tasks": {
+    "cross_modal_retrieval": {
+      "metric_key": "cross_modal_retrieval_mrr",
+      "metrics_json": "results/omni_finetune/xperience10m_qwen3_omni_v6_cross_modal_retrieval_probe_a100_20260618T000000Z/cross_modal_retrieval/metrics.json",
+      "num_samples": 4032,
+      "primary_score": 0.5080191798941732,
+      "task_label": "Cross-Modal Retrieval",
+      "task_number": 9
+    }
+  },
+  "title": "Qwen3-Omni v6 Retrieval Task Probes"
+}

results/omni_finetune/xperience10m_qwen3_omni_v6_retrieval_task_probes_a100_20260617T175919Z/RUN_REPORT.md

@@ -0,0 +1,8 @@
+# Qwen3-Omni v6 Retrieval Task Probes
+
+- Run ID: `xperience10m_qwen3_omni_v6_retrieval_task_probes_a100_20260617T175919Z`
+- Shards: `2`
+
+| Task | Metric | Score | Samples |
+| --- | --- | ---: | ---: |
+| Language Grounding | caption_grounding_mrr | 0.876447 | 4032 |

results/omni_finetune/xperience10m_qwen3_omni_v6_retrieval_task_probes_a100_20260617T175919Z/caption_grounding/metrics.json

@@ -0,0 +1,24 @@
+{
+  "adapter_dir": "/mnt/kgc/chaoyue/ropedia-h20-side/ropedia-episode-task-suite/checkpoints/xperience10m_qwen3_omni_128ep_multiscale_cap96_v6_rank64_lr5e5_full8gpu_lora/adapter_lora",
+  "candidate_count": 4,
+  "caption_grounding_mrr": 0.8764467592592605,
+  "dataset_jsonl": "/mnt/kgc/chaoyue/ropedia-h20-side/ropedia-episode-task-suite/results/omni_finetune/xperience10m_qwen3_omni_128ep_multiscale_cap96_v5_full8gpu_lora_dataset/dataset_a100_eval.jsonl",
+  "eval_split": "test",
+  "metric_key": "caption_grounding_mrr",
+  "model_id": "/mnt/kgc/chaoyue/ropedia-h20-side/modelscope_models/Qwen__Qwen3-Omni-30B-A3B-Instruct",
+  "mrr": 0.8764467592592605,
+  "num_samples": 4032,
+  "primary_metric": "caption_grounding_mrr",
+  "primary_score": 0.8764467592592605,
+  "run_id": "xperience10m_qwen3_omni_v6_retrieval_task_probes_a100_20260617T175919Z",
+  "sample_offset": 0,
+  "sample_stride": 1,
+  "scope": "held_out_test_qwen3_retrieval_task_probe",
+  "score_policy": "GPU-backed Qwen3-Omni v6 text-to-video retrieval probe. The text query is built from held-out action/subtask/object labels, candidates are shuffled staged mosaic video windows, and the score is MRR of the true window. This does not score tasks whose numeric/raw targets are absent from the export.",
+  "status": "pass",
+  "task_id": "caption_grounding",
+  "task_label": "Language Grounding",
+  "task_number": 8,
+  "title": "Qwen3-Omni v6 Language Grounding",
+  "top1_accuracy": 0.7864583333333334
+}

results/omni_finetune/xperience10m_qwen3_omni_v6_retrieval_task_probes_a100_20260617T175919Z/collection_validation.json

@@ -0,0 +1,16 @@
+{
+  "records": [
+    {
+      "metric_key": "caption_grounding_mrr",
+      "num_samples": 4032,
+      "primary_score": 0.8764467592592605,
+      "source": "results/omni_finetune/xperience10m_qwen3_omni_v6_retrieval_task_probes_a100_20260617T175919Z/caption_grounding/metrics.json",
+      "task_id": "caption_grounding"
+    }
+  ],
+  "run_id": "xperience10m_qwen3_omni_v6_retrieval_task_probes_a100_20260617T175919Z",
+  "status": "pass",
+  "summary": "results/omni_finetune/xperience10m_qwen3_omni_v6_retrieval_task_probes_a100_20260617T175919Z/summary.json",
+  "title": "Qwen3 Future Task Probe Collection Validation",
+  "validated_task_count": 1
+}

results/omni_finetune/xperience10m_qwen3_omni_v6_retrieval_task_probes_a100_20260617T175919Z/launch_env.txt

@@ -0,0 +1,11 @@
+run_id=xperience10m_qwen3_omni_v6_retrieval_task_probes_a100_20260617T175919Z
+dataset_jsonl=/mnt/kgc/chaoyue/ropedia-h20-side/ropedia-episode-task-suite/results/omni_finetune/xperience10m_qwen3_omni_128ep_multiscale_cap96_v5_full8gpu_lora_dataset/dataset_a100_eval.jsonl
+model_dir=/mnt/kgc/chaoyue/ropedia-h20-side/modelscope_models/Qwen__Qwen3-Omni-30B-A3B-Instruct
+adapter_dir=/mnt/kgc/chaoyue/ropedia-h20-side/ropedia-episode-task-suite/checkpoints/xperience10m_qwen3_omni_128ep_multiscale_cap96_v6_rank64_lr5e5_full8gpu_lora/adapter_lora
+tasks=caption_grounding
+candidate_count=4
+cuda_device_groups=0,1 2,3
+shards=2
+started_at=2026-06-18T05:20:16+08:00
+finished_at=2026-06-18T08:41:05+08:00
+exit_code=0

results/omni_finetune/xperience10m_qwen3_omni_v6_retrieval_task_probes_a100_20260617T175919Z/summary.json

@@ -0,0 +1,19 @@
+{
+  "run_id": "xperience10m_qwen3_omni_v6_retrieval_task_probes_a100_20260617T175919Z",
+  "shard_dirs": [
+    "results/omni_finetune/xperience10m_qwen3_omni_v6_retrieval_task_probes_a100_20260617T175919Z_shard0",
+    "results/omni_finetune/xperience10m_qwen3_omni_v6_retrieval_task_probes_a100_20260617T175919Z_shard1"
+  ],
+  "status": "pass",
+  "tasks": {
+    "caption_grounding": {
+      "metric_key": "caption_grounding_mrr",
+      "metrics_json": "results/omni_finetune/xperience10m_qwen3_omni_v6_retrieval_task_probes_a100_20260617T175919Z/caption_grounding/metrics.json",
+      "num_samples": 4032,
+      "primary_score": 0.8764467592592605,
+      "task_label": "Language Grounding",
+      "task_number": 8
+    }
+  },
+  "title": "Qwen3-Omni v6 Retrieval Task Probes"
+}