- Rating
- 2
- Confidence
- 3
- Informativeness
- 2
- Presentation
- 3
- Soundness
- 2
Read review
Paper summary
CoDance trains the reasoner for prompt enhancement and the video diffusion model for generation together, for the improvement of video physical plausibility. CoDanceRL uses shared nested rollouts, physical fidelity reward and geometric trajectory reward to update both components. The model performs well on Physics-IQ Verified, RBench and WorldArena.
Critical strengths
- Improvements on the held-out VLM judge and downstream planning can prove that training produces useful changes rather than only improving motion plausibility.
- Ablations on coupled training (compared with reasoner-only, generator-only, sequential) are compelling.
Critical weaknesses
- CoDance only reports 2 tasks of WorldArena, which is hard to prove its general advantage
- Metrics of Physics-IQ are used as reward which is very unfair for comparison.
- RBench covers a very small range of models to compare.
Questions for authors
- Does the coupled/sequential advantage persist across training seeds since the distinction is very small?
- Which exact RBench judge settings were used, and were the imported baseline scores obtained with a compatible evaluator?
- What are the performances of CoDance on the whole WorldArena benchmark?