| --- |
| license: apache-2.0 |
| task_categories: |
| - video-classification |
| language: |
| - en |
| tags: |
| - reward-model |
| - world-model |
| - video-generation |
| - camera-control |
| - preference-data |
| size_categories: |
| - n<1K |
| configs: |
| - config_name: default |
| data_files: bench.jsonl |
| --- |
| |
| # WorldReward-Bench |
|
|
| A human-annotated preference benchmark for **camera-conditioned world models**. |
| 760 pairs of videos, each pair generated by two different models from the *same* |
| source image and the *same* camera-action sequence, with human verdicts on three |
| independent axes. |
|
|
| - 📰 Paper: https://arxiv.org/abs/2609.03952 |
| - 🪐 Project Page: https://codegoat24.github.io/WorldReward |
| - 🤗 Model Collections: https://huggingface.co/CodeGoat24/WorldReward-9B |
| - 🚀 Github: https://github.com/CodeGoat24/WorldReward |
| - 👋 Point of Contact: [Yibin Wang](https://codegoat24.github.io) |
|
|
|  |
|
|
|
|
| ## Evaluation Dimensions |
|
|
| | Axis | Question | |
| |---|---| |
| | `action` | Did the camera actually execute the commanded motion — right direction, right magnitude, no un-commanded drift? | |
| | `appearance` | Which video looks better — fewer artifacts, more stable structure, faithful to the source scene? | |
| | `motion` | Which video is genuinely *generating* new content, rather than sliding a static texture or melting? | |
|
|
| Each verdict is `left`, `right`, or `tie`. |
|
|
| ## Contents |
|
|
| ``` |
| bench.jsonl # 760 pairs, one JSON object per line |
| videos/<pair_id>/source.* # shared source image (exact path is in bench.jsonl) |
| videos/<pair_id>/left.mp4 # the video shown on the left |
| videos/<pair_id>/right.mp4 # the video shown on the right |
| videos/<pair_id>/left_overlay.mp4 # left.mp4 with the commanded action burned in |
| videos/<pair_id>/right_overlay.mp4 # right.mp4 with the commanded action burned in |
| ``` |
|
|
|
|
| The `*_overlay.mp4` files are a visualisation aid for inspecting trajectories by |
| eye. |
|
|
| Source images keep their original format and pixels when they are at most |
| 2048px on the long side. Larger ones are downscaled to 2048px and saved |
| as JPEG. |
|
|
| ## Schema |
|
|
| | Field | Type | Description | |
| |---|---|---| |
| | `pair_id` | string | Opaque identifier, `wrb_0001`-style. Carries no metadata. | |
| | `input_image` | path | Source image both videos were generated from | |
| | `input_caption` | string | English description of the source scene | |
| | `actions` | list[string] | Commanded camera-action sequence, one token per step | |
| | `frames_per_action` | int | Frames each action occupies | |
| | `num_frames` | int | Total frames per video | |
| | `left` / `right` | object | `video`, `overlay`, and the generating `model` name | |
| | `trajectory_family` | string | One of 9 fine-grained trajectory types | |
| | `trajectory_group` | string | `pure_translation` / `pure_rotation` / `compound` | |
| | `style` | string | `photo` / `game_anime` / `traditional` | |
| | `label` | object | Human verdicts: `action`, `appearance`, `motion`, each `left`/`right`/`tie` | |
|
|
|
|
|
|
| ## Results |
|
|
| Three-way agreement with the human labels (%). All 760 pairs count; a pair whose |
| label is `tie` is correct only if the model also predicts `tie`. Act./App./Mot. = |
| action / appearance / motion. **Best** and _second-best_ per column; `--` marks an |
| axis a predictor does not model. |
|
|
| | Reward model | All<br>Act. | All<br>App. | All<br>Mot. | Translation<br>Act. | Translation<br>App. | Translation<br>Mot. | Rotation<br>Act. | Rotation<br>App. | Rotation<br>Mot. | Compound<br>Act. | Compound<br>App. | Compound<br>Mot. | Photo<br>Act. | Photo<br>App. | Photo<br>Mot. | Game/Anime<br>Act. | Game/Anime<br>App. | Game/Anime<br>Mot. | Art<br>Act. | Art<br>App. | Art<br>Mot. | |
| |:---|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:| |
| | *Closed-source VLM* | | | | | | | | | | | | | | | | | | | | | | |
| | Gemini-3.1-Pro | 65.79 | _80.13_ | 60.79 | 64.73 | **82.19** | 64.38 | 62.90 | 83.87 | 57.53 | 68.79 | _75.53_ | 59.22 | 65.48 | **82.74** | 63.84 | 65.27 | 79.34 | 57.49 | 70.49 | 68.85 | 60.66 | |
| | GPT-5.5 | _74.21_ | 79.87 | _69.47_ | _72.95_ | _81.16_ | _74.66_ | _72.04_ | _84.41_ | 62.90 | _76.95_ | 75.53 | _68.44_ | _76.44_ | _81.64_ | _68.77_ | 71.56 | _79.64_ | _68.86_ | **75.41** | _70.49_ | **77.05** | |
| | *Image / video quality reward models* | | | | | | | | | | | | | | | | | | | | | | |
| | VideoAlign | -- | 61.32 | 40.13 | -- | 66.44 | 31.51 | -- | 60.22 | 45.16 | -- | 56.74 | 45.74 | -- | 61.10 | 41.10 | -- | 61.08 | 41.32 | -- | 63.93 | 27.87 | |
| | UnifiedReward-Flex | -- | 64.62 | 49.86 | -- | 63.90 | 46.57 | -- | 73.18 | 56.98 | -- | 59.78 | 48.55 | -- | 65.08 | 55.31 | -- | 63.64 | 47.02 | -- | 67.27 | 30.91 | |
| | UnifiedReward-Think | -- | 66.09 | 38.79 | -- | 65.41 | 32.88 | -- | 69.73 | 52.43 | -- | 64.41 | 35.94 | -- | 69.51 | 41.21 | -- | 64.37 | 37.43 | -- | 55.00 | 31.67 | |
| | Aesthetic | -- | 69.87 | -- | -- | 66.10 | -- | -- | 72.04 | -- | -- | 72.34 | -- | -- | 66.58 | -- | -- | 75.15 | -- | -- | 60.66 | -- | |
| | HPSv3 | -- | 73.68 | -- | -- | 74.66 | -- | -- | 76.34 | -- | -- | 70.92 | -- | -- | 73.15 | -- | -- | 75.45 | -- | -- | 67.21 | -- | |
| | *Geometry estimation models* | | | | | | | | | | | | | | | | | | | | | | |
| | DAv3 | 70.53 | -- | -- | 67.47 | -- | -- | 68.82 | -- | -- | 74.82 | -- | -- | 70.96 | -- | -- | _75.75_ | -- | -- | 39.34 | -- | -- | |
| | WorldMirror | 68.55 | -- | -- | 67.81 | -- | -- | 68.28 | -- | -- | 69.50 | -- | -- | 67.40 | -- | -- | 74.25 | -- | -- | 44.26 | -- | -- | |
| | *Backbone, zero-shot* | | | | | | | | | | | | | | | | | | | | | | |
| | Qwen3.5-VL-9B | 48.42 | 48.29 | 43.82 | 48.29 | 45.55 | 41.78 | 52.69 | 52.15 | 49.46 | 45.74 | 48.58 | 42.20 | 50.14 | 43.84 | 47.12 | 47.01 | 52.10 | 42.81 | 45.90 | 54.10 | 29.51 | |
| | Qwen3.5-VL-27B | 63.68 | 44.34 | 62.76 | 65.07 | 37.33 | 65.75 | 65.05 | 51.61 | _63.44_ | 61.35 | 46.81 | 59.22 | 64.93 | 38.36 | 66.85 | 62.87 | 51.50 | 58.68 | 60.66 | 40.98 | 60.66 | |
| | **WorldReward-9B** | **77.63** | **81.32** | **73.03** | **76.71** | 77.74 | **78.77** | **73.12** | **86.02** | **64.52** | **81.56** | **81.91** | **72.70** | **77.26** | 81.37 | **71.78** | **78.74** | **82.04** | **75.75** | _73.77_ | **77.05** | _65.57_ | |
|
|
|
|
| ## Usage |
| See https://github.com/CodeGoat24/WorldReward for the full evaluation protocol. |
|
|
|
|
| ## Citation |
|
|
| ```bibtex |
| @article{wang2026worldreward, |
| title={WorldReward: Reward Modeling for Camera-Conditioned World Models}, |
| author={Wang, Yibin and Wang, Zehan and Tang, Junshu and Li, Zhimin and Zhou, Yujie and Bu, Jiazi and Ling, Pengyang and Han, Feng and Zhang, Zhixiong and Xing, Long and others}, |
| journal={arXiv preprint arXiv:2609.03952}, |
| year={2026} |
| } |
| ``` |
|
|
