目的地之前：视觉忠实性在慢思考中的重要性
Journey Before Destination: On the importance of Visual Faithfulness in Slow Thinking
AI总结：
本文提出视觉忠实性评估方法，通过检测并修正不忠实的感知步骤提升多模态推理的可靠性。
摘要：
增强推理的视觉语言模型（VLMs）生成显式的推理链，承诺更大的能力和透明度，但也引入了新的失败模式：模型可能通过视觉不忠实的中间步骤达到正确答案，或者推理忠实但最终预测失败。标准评估仅测量最终答案的准确性无法区分这些行为。我们引入推理链的视觉忠实性作为独特的评估维度，关注推理链的感知步骤是否基于图像。我们提出一种训练和参考免费的框架，将链条分解为感知与推理步骤，并使用现成的VLM判断者进行步骤级的忠实性评估，此外通过人类元评估验证这种方法。基于此指标，我们提出一种轻量的自我反思程序，能够检测并局部重新生成不忠实的感知步骤而无需任何训练。在多个推理训练的VLM和感知密集的基准测试中，我们的方法降低了不忠实感知率，同时保持最终答案的准确性，提高了多模态推理的可靠性。
英文摘要：
Reasoning-augmented vision language models (VLMs) generate explicit chains of thought that promise greater capability and transparency but also introduce new failure modes: models may reach correct answers via visually unfaithful intermediate steps, or reason faithfully yet fail on the final prediction. Standard evaluations that only measure final-answer accuracy cannot distinguish these behaviors. We introduce the visual faithfulness of reasoning chains as a distinct evaluation dimension, focusing on whether the perception steps of a reasoning chain are grounded in the image. We propose a training- and reference-free framework that decomposes chains into perception versus reasoning steps and uses off-the-shelf VLM judges for step-level faithfulness, additionally verifying this approach through a human meta-evaluation. Building on this metric, we present a lightweight self-reflection procedure that detects and locally regenerates unfaithful perception steps without any training. Across multiple reasoning-trained VLMs and perception-heavy benchmarks, our method reduces Unfaithful Perception Rate while preserving final-answer accuracy, improving the reliability of multimodal reasoning.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
