跳到正文
arXiv cs.AI· Yu Yuan, Ruizhou Ding, Shaoheng Fang, Stanley H. Chan, Zhuo Chen·· 10 小时前AI 评分38

EVA-RSI:视频评估器的递归自我改进,用于评判时序质量

EVA-RSI: Recursive Self-Improvement of Video Evaluators for Judging Temporal Quality

AI 导读

EVA-RSI 是一个多智能体框架,在无验证器、多维度同时评判的约束下探索视频评估器的递归自我改进。它通过时间反转、丢帧等受控编辑构建固定代理考试,无需人工标注即可提供监督,并以少量人工标注真实视频锚定人类判断。从 Qwen3-VL-8B 起步,该方法提升了运动评估,平滑度增益最大,且在公开基准的 dynamics 子集上取得 21 个测试模型中的最高分。

正文 · 原文

View PDF HTML (experimental)

Abstract:Video evaluators play an important role in video generation, yet they miss motion defects, limiting the quality of generated videos. Recursive self-improvement (RSI) has made strong progress in mathematics and coding, where explicit verifiers provide reliable feedback on a single goal. Video evaluation poses two difficulties for RSI: it has no verifier, and it must judge many dimensions at once, such as physics, smoothness and dynamics, where a gain on one can cost another. We introduce EVA-RSI, a multi-agent framework for exploring RSI under these two constraints. We build a fixed proxy exam with construction-defined targets by applying controlled changes to videos, such as temporal reversal and frame dropping. These edits supply supervision without human annotation. A small set of human-labeled real videos anchors the exam to human judgment. Each dimension has an LLM agent that proposes improvements. A leader agent coordinates the proposals, and a Nash bargaining rule selects updates that balance progress across dimensions. Humans design the exam, review candidate changes and revise the curriculum between stages. Starting from Qwen3-VL-8B, EVA-RSI improves motion evaluation, with the largest gain on smoothness, and the gains carry over to defect types never seen during the run. On the dynamics subsets of public benchmarks, our evaluator achieves the highest observed scores among 21 tested models. As an RL reward, it increases generated motion and earns human preference. A data-selection pipeline incorporating it also increases motion.
Subjects: Artificial Intelligence (cs.AI)
Cite as: arXiv:2611.23890 [cs.AI]
  (or arXiv:2611.23890v1 [cs.AI] for this version)
  https://doi.org/10.48550/arXiv.2611.23890

arXiv-issued DOI via DataCite (pending registration)

Submission history

From: Yu Yuan [view email]
[v1] Wed, 18 Nov 2026 05:41:49 UTC (4,132 KB)

来源:arXiv cs.AI · arxiv.org