Datasets:

Video-Reason
/

VBVR-Pro-RL

@@ -1,113 +0,0 @@
----
-language:
-- en
-tags:
-- video-generation
-- image-to-video
-- visual-reasoning
-- reinforcement-learning
-pretty_name: VBVR-Pro-RL
-size_categories:
-- 10K<n<100K
----
-# VBVR-Pro-RL
-The **reinforcement-learning split** of VBVR-Pro: 50 parameterized tasks × 1,000 instances, held out from the SFT splits, in both a video (I2V) and an interleaved-image setting.
-Companion resources: [VBVR-Pro-SFT-Video](https://huggingface.co/datasets/Video-Reason/VBVR-Pro-SFT-Video) · [VBVR-Pro-SFT-Image](https://huggingface.co/datasets/Video-Reason/VBVR-Pro-SFT-Image) · [VBVR-Pro-Bench](https://huggingface.co/datasets/Video-Reason/VBVR-Pro-Bench)
-## At a glance
-| Property | Value |
-|---|---|
-| Tasks | **50** |
-| Instances per task | **1,000** |
-| Total instances | **50,000** per setting |
-| Archives | 50 video + 50 image tar.gz |
-| Total size | **10.9 GB** (video 9.7 + image 1.2) |
-| Resolution | **1024 × 1024**, as generated |
-| Video | 16 fps, MPEG-4 Part 2, no re-encode |
-## Layout
-```
-.
-├── VBVR-Pro-RL-Video/
-│   ├── G-131_select_next_figure_increasing_size_sequence_data-generator.tar.gz
-│   └── …                                                     # 50 archives
-├── VBVR-Pro-RL-Image/
-│   ├── G-131_select_next_figure_increasing_size_sequence_data-generator.tar.gz
-│   └── …                                                     # 50 archives
-├── video_jsonl/                                              # 50 files
-├── image_jsonl/                                              # 50 files
-├── meta_video_rl.json
-└── meta_image_rl.json
-```
-An archive under `VBVR-Pro-RL-Video/` extracts to:
-```
-G-131_select_next_figure_increasing_size_sequence_data-generator/        # task
-└── select_next_figure_increasing_size_sequence_task/                    # subtask
-    ├── select_next_figure_increasing_size_sequence_00005005/            # sample
-    │   ├── first_frame.png          # conditioning image  (1024 × 1024)
-    │   ├── metadata.json            # task parameters, ground truth, scoring contract
-    │   └── video/
-    │       ├── prompt.txt           # instruction
-    │       ├── ground_truth.mp4     # reference video     (16 fps, MPEG-4 Part 2)
-    │       └── final_frame.png      # last frame of the reference video
-    ├── select_next_figure_increasing_size_sequence_00005006/            # same files
-    ├── …
-    └── select_next_figure_increasing_size_sequence_00006004/            # 1,000 samples per task
-```
-The matching archive under `VBVR-Pro-RL-Image/` has the same task, subtask and
-sample names, with the reference output as an image sequence instead:
-```
-G-131_select_next_figure_increasing_size_sequence_data-generator/
-└── select_next_figure_increasing_size_sequence_task/
-    ├── select_next_figure_increasing_size_sequence_00005005/
-    │   ├── first_frame.png          # conditioning image  (1024 × 1024)
-    │   ├── metadata.json            # task parameters, ground truth, scoring contract
-    │   └── image/
-    │       ├── prompt.txt           # instruction
-    │       ├── frame_1.png          # reference output, step 1
-    │       └── …                    # up to frame_N.png
-    ├── select_next_figure_increasing_size_sequence_00005006/
-    ├── …
-    └── select_next_figure_increasing_size_sequence_00006004/
-```
-Sample ids run `00005005`–`00006004`, disjoint from the SFT splits
-(`00000000`–`00004999`) by construction. `N` is task-dependent: the number of
-steps the instruction asks the model to render, stated in `prompt.txt`.
-## Usage
-```bash
-huggingface-cli download Video-Reason/VBVR-Pro-RL --repo-type dataset --local-dir ./rl
-mkdir -p rl-data && for f in ./rl/VBVR-Pro-RL-Video/*.tar.gz; do tar xzf "$f" -C rl-data; done
-```
-### Index files
-`meta_video_rl.json` and `meta_image_rl.json` map every task to its jsonl under
-`video_jsonl/` and `image_jsonl/`. `root` is the directory the archives were
-extracted into and `length` is that task's row count (1,000 throughout).
-Video rows carry `sample_id`, `first_frame`, `metadata`, `clip_path`,
-`final_frame` and `prompt`, every path relative to `root` — the same schema as
-[VBVR-Pro-SFT-Video](https://huggingface.co/datasets/Video-Reason/VBVR-Pro-SFT-Video).
-Image rows carry `id`, `image`, `conversations` and `annotation_steps`, matching
-[VBVR-Pro-SFT-Image](https://huggingface.co/datasets/Video-Reason/VBVR-Pro-SFT-Image).
-> **The image split's reasoning text is a fixed placeholder.** Its images,
-> prompts and ground truth are real, but the `<think>` sentences in
-> `conversations` and `annotation_steps` come from a template — no annotated
-> reasoning traces exist for this split. Do not train a reasoning-text objective
-> on them, and do not report them as annotated chain-of-thought.
-> Extract this split into a **separate** directory from the SFT splits. They share task names but come from different generator runs. Sample ids do not collide — RL instances are numbered `00005005`–`00006004`, SFT `00000000`–`00004999` — but the two trees are not interchangeable.