研究人员开发了感知流网络（PFlowNet）以提高大型视觉语言模型（LVLMs）的视觉推理能力。PFlowNet将感知与推理分离，并使用变分强化学习来指导感知行为，旨在减少语言偏见和幻觉。该方法在V* Bench和MME-RealWorld-lite等基准测试中取得了最先进的成果。另一个相关模型VGR通过将语言推断基础化到检测到的图像区域中来增强多模态推理能力，在ChartQA等基准测试中显示出显著的改进，同时使用的图像令牌更少。 AI
影响 引入了多模态推理的新型架构，有望提高视觉-语言模型的准确性并减少幻觉。
排序理由 该集群包含两篇arXiv论文，详细介绍了用于视觉推理的新模型和方法。
AI 生成摘要 · Google Gemini · 来自 3 个来源。 我们如何撰写摘要 →
arXiv:2605.02730v1 Announce Type: new Abstract: Despite the success of Large-Vision Language Models (LVLMs), general optimization objectives (e.g., standard MLE) fail to constrain visual trajectories, leading to language bias and hallucination. To mitigate this, current methods i…
Despite the success of Large-Vision Language Models (LVLMs), general optimization objectives (e.g., standard MLE) fail to constrain visual trajectories, leading to language bias and hallucination. To mitigate this, current methods introduce geometric priors from visual experts as…
arXiv:2506.11991v3 Announce Type: replace Abstract: In the field of multimodal chain-of-thought (CoT) reasoning, existing approaches predominantly rely on reasoning on pure language space, which inherently suffers from language bias and is largely confined to math or science doma…