通过感知扰动和奖励建模减轻多模态大语言模型作为评判者中的感知判断偏差
Mitigating Perceptual Judgment Bias in Multimodal LLM-as-a-Judge via Perceptual Perturbation and Reward Modeling
AI总结：
本文通过构建感知扰动数据集和结合GRPO奖励与批排序目标的统一训练框架，解决了多模态大语言模型作为评判者时因视觉证据与文本线索冲突而产生的感知判断偏差问题，显著提升了感知忠实度和与人类评价的一致性。
摘要：
最近的多模态大语言模型展示了强大的推理能力，但它们作为自动评估器的可靠性仍然受到一个关键弱点的限制：当视觉证据与文本线索冲突时，多模态大语言模型评判者倾向于奖励看似合理的叙述而非感知上正确的答案。我们识别并系统分析了这一现象，称之为感知判断偏差。通过受控的视觉扰动，现有的多模态评判者经常锚定于响应文本而非自身的视觉感知，导致不一致且不可验证的评估。为了解决这个问题，我们引入了感知扰动判断数据集，该数据集构建了最小编辑的反事实响应，隔离了感知错误并实现了可验证的监督。基于该数据集，我们开发了一个统一的训练框架，将结构化的基于GRPO的奖励与批排序目标相结合，实现了无需显式成对标签的连贯全局排序。在多种多模态大语言模型作为评判者的基准测试上的实验表明，我们的方法显著提高了感知忠实度、排序连贯性以及与人类评价的一致性。我们的结果为训练感知基础、可解释且对视觉推理冲突鲁棒的多模态评判者建立了一条可扩展且可泛化的路径。
英文摘要：
Recent multimodal large language models have demonstrated strong reasoning ability, yet their reliability as automated evaluators remains limited by a critical weakness: when visual evidence conflicts with textual cues, MLLM judges tend to reward plausible narratives over perceptually correct answers. We identify and systematically analyze this phenomenon, which we term Perceptual Judgment Bias. Through controlled visual perturbations, existing multimodal judges frequently anchor on the response text instead of their own visual perception, leading to inconsistent and non-verifiable evaluations. To address this issue, we introduce the Perceptually Perturbed Judgment Dataset, which constructs minimally edited counterfactual responses that isolate perceptual errors and enable verifiable supervision. Building on this dataset, we develop a unified training framework that combines a structured GRPO-based reward with a batch-ranking objective, achieving coherent global ordering without explicit pairwise labels. Experiments across diverse MLLM-as-a-Judge benchmarks show that our approach substantially improves perceptual fidelity, ranking coherence, and alignment with human evaluation. Our results establish a scalable and generalizable pathway for training multimodal judges that are perceptually grounded, interpretable, and robust to visual-reasoning conflicts.
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。
