New reward model uses visual artifacts to improve video generation quality

Researchers have developed ArtiVidRM, a novel reward modeling approach for video generation that explicitly identifies and uses visual artifacts as rubrics for quality assessment. This method, which follows an observe-then-score paradigm, organizes visible generation artifacts into a structured list to ground each score in concrete evidence, making the reward more interpretable and informative for optimization. Experiments demonstrate that ArtiVidRM achieves state-of-the-art performance on video preference benchmarks and can guide the refinement of video generators. AI

IMPACT This new reward modeling technique could lead to more interpretable and controllable video generation systems.

RANK_REASON The cluster contains a research paper detailing a new methodology for video generation. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CV →

• alphaXiv
• ArtiVidRM
• arXiv
• CatalyzeX Code Finder for Papers
• Connected Papers
• CORE Recommender
• DagsHub
• Gotit.pub
• Hugging Face
• Litmaps
• ScienceCast
• scite Smart Citations

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →