paper-with-me

Papers

Mitigating Perceptual Judgment Bias in Multimodal LLM-as-a-Judge via Perceptual Perturbation and Reward Modeling

2026-06-01 · Seojeong Park, Jiho Choi, Junyong Kang, Seonho Lee, Jaeyo Shin, Hyunjung Shim arxiv

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.

📄 PDF Abstract BibTeX arXiv:2606.02578

Code (0)

등록된 구현이 없습니다.

Similar Papers 제목 키워드 기반

Calibrating MLLM-as-a-judge via Multimodal Bayesian Prompt Ensembles

2025-09-10 · Eric Slyman, Mehrab Tanjim, Kushal Kafle, Stefan Lee arxiv

Multimodal large language models (MLLMs) are increasingly used to evaluate text-to-image (TTI) generation systems, providing automated judgments based on visual and textual context. However, these "judge" models often su…

Image Clustering

MJ1: Multimodal Judgment via Grounded Verification

2026-03-09 · Bhavesh Kumar, Dylan Feng, Leonard Tang arxiv

Multimodal judges struggle to ground decisions in visual evidence. We present MJ1, a multimodal judge trained with reinforcement learning that enforces visual grounding through a structured grounded verification chain (o…

Reinforcement LearningMultimodal ReasoningVisual GroundingImage Editing

Toward Robust LLM-Based Judges: Taxonomic Bias Evaluation and Debiasing Optimization

2026-03-09 · Hongli Zhou, Hui Huang, Rui Zhang, Kehai Chen 외 arxiv

Large language model (LLM)-based judges are widely adopted for automated evaluation and reward modeling, yet their judgments are often affected by judgment biases. Accurately evaluating these biases is essential for ensu…

Reinforcement LearningContrastive Learning

Evaluating and Mitigating LLM-as-a-judge Bias in Communication Systems

2025-10-14 · Jiaxin Gao, Chen Chen, Yanwen Jia, Xueluan Gong 외 arxiv

Large Language Models (LLMs) are increasingly being used to autonomously evaluate the quality of content in communication systems, e.g., to assess responses in telecom customer support chatbots. However, the impartiality…

Mitigating Judgment Preference Bias in Large Language Models through Group-Based Polling

2025-10-09 · Shuliang Liu, Zhipeng Xu, Zhenghao Liu, Yukun Yan 외 arxiv

Large Language Models (LLMs) as automatic evaluators, commonly referred to as LLM-as-a-Judge, have also attracted growing attention. This approach plays a vital role in aligning LLMs with human judgments, providing accur…