VA-Judger: Reward Modeling from Human Preference Feedback for Joint Video-Audio Generation
- 类型：arxiv
- 标识：2608.18607
- 链接：https://arxiv.org/abs/2608.18607
- 主分类：multimodal
- 形态：method
- 被引：0
- 被引来源：Semantic Scholar
- S2被引：0
- 影响力被引：0
- TLDR：VA-Judger, a chain-of-thought omni-reward model for joint video-audio generation that first learns from pairs with clear quality gaps to establish structured output and coarse preference discrimination, then distills reliable preference explanations for harder near-quality comparisons via rejection sampling verified against human annotations, and finally performs dimension-wise reinforcement learning that decomposes human feedback into individual quality dimensions for denser reward signals.
- 待LLM分类：否
- 标题中文：VA-Judger：基于人类偏好反馈的联合音视频生成奖励建模
- TLDR中文：VA-Judger：用于联合视频-音频生成的思维链全模态奖励模型。它首先从具有明显质量差距的数据对中学习，以建立结构化输出与粗粒度偏好判别；然后通过对照人工标注进行拒绝采样，蒸馏得到针对更难近质量对比的可靠偏好解释；最终执行按维度分解的强化学习，将人类反馈分解为各个质量维度以获得更密集的奖励信号。
- 来源文件：
- /inbox/tom/_candidates/2026-08-21-agent-rag-longcontext-candidates.json
- [S2 enrich]