JiaxinGe's picture
Upload README.md with huggingface_hub
a5cab47 verified
|
Raw
History Blame
621 Bytes
---
license: cc-by-nc-4.0
tags: [bagel, vlm-gym, world-model, maze2d, dataset]
---
# maze2d_easy_cot_chunk_k3_train
BAGEL VLM-Gym world-model dataset (maze2d / cot).
- CoT chunk-K train set: all-step interleaved imagined reasoning; re-grounds on the true frame every K=3 steps.
- layout: Train-only. Gzipped-JSONL shards under `training/`; each row is one packed SFT sample with base64-JPEG frames inline.
- images are base64-encoded JPEG frames stored inline in each JSONL row.
Pairs with the matching `maze2d` checkpoint(s) under the companion model org; CoT and non-CoT
variants share the same underlying rollouts.