Research map / Video / Video generation
Thinking with Frames: Generative Video Distortion Evaluation via Frame Reward Model
Yuan Wang, Borui Liao, Huijuan Huang, Jinda Lu, Ouxiang Li, Kuien Liu, Meng Wang, Xiang Wang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add SFT before GRPO (ablation) | REACT · REACT-Video | accuracy w/ tie: 0.387 -> 0.610 | Improved |
| use BTT-based preference reward instead of binary reward (ablation) | REACT GRPO · REACT-Video | accuracy w/ tie: 0.352 -> 0.610 | Improved |
| add dynamic frame sampling (ablation) | REACT · REACT-Video | accuracy w/o tie: 0.725 -> 0.813 | Improved |
| add GRPO after SFT (ablation) | REACT · REACT-Frame distorted | F1-score: 0.764 -> 0.845 | Improved |
| add masked loss in second SFT epoch (ablation) | SFT · REACT-Frame distorted | F1-score: 0.690 -> 0.764 | Improved |
| train SFT for 2 epochs instead of 1 (ablation) | SFT · REACT-Frame distorted | F1-score: 0.557 -> 0.690 | Improved |
| SFT then GRPO on normal frames vs masked SFT (ablation) | REACT · REACT-Frame normal | recall: 0.659 -> 0.594 | Hurt |
| use frame-level REACT instead of UnifiedReward | preference alignment · REACT-Video | accuracy w/o tie: 0.701 -> 0.813 | Improved |
| use REACT instead of general MLLM Gemini-2.5-Pro | distortion recognition · REACT-Frame distorted | F1-score: 0.650 -> 0.845 | Improved |
Closest papers by meaning
- Improving Video Generation with Human Feedback (NeurIPS 2025 · similarity 0.88)
- InstructVideo: Instructing Video Diffusion Models with Human Feedback (CVPR 2024 · similarity 0.88)
- Think, then Score: Decoupled Reasoning and Scoring for Video Reward Modeling (NeurIPS 2026 · similarity 0.87)
- Video-Bench: Human-Aligned Video Generation Benchmark (CVPR 2025 · similarity 0.87)
- VBench: Comprehensive Benchmark Suite for Video Generative Models (CVPR 2024 · similarity 0.87)
- EvalCrafter: Benchmarking and Evaluating Large Video Generation Models (CVPR 2024 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
