AI.news
主页教程研究工具模型AI创业讨论新闻每日简报WIKI🚀 创业库★ 投稿
AI+医疗机器人教育金融能源健康娱乐思考

Improving Multimodal Reasoning via Worst Dimension Optimization

arxiv.org
分享

View PDF HTML (experimental)

Abstract:Multimodal reasoning requires a path that retains integrity over a wide range of constraints, from visual grounding to logic consistency. However, the current Process Reward Models focus on heuristically defined rewards that equally weigh these factors, which may lead to the concealment of individual dimension failures by the dominating factors, without guaranteeing the validity of the reasoning process in general.

Submission history

From: Haocheng Lv [view email]
[v1] Fri, 5 Jun 2026 19:32:23 UTC (1,088 KB)