When Vision-Language Models Judge Without Seeing: Exposing Informativeness Bias
Xiaohan Zou, Roshan Sridhar, Mohammadtaher Safarzadeh, Dan Roth
Abstract
The reliability of VLM-as-a-Judge is critical for the automatic evaluation of vision-language models (VLMs). Despite recent progress, our analysis reveals that VLM-as-a-Judge often pays limited attention to the image when making decisions. Instead, they often blindly favor the more informative answer, even when they can recognize it conflicts with the image content. We call this problem informativeness bias, which significantly undermines judge reliability. To address it, we propose BIRCH (Balanced Informativeness and CoRrectness with a Truthful AnCHor), a judging paradigm that first corrects inconsistencies with the image content in candidate answers, and then compares the answers against this corrected version. This shifts the judge's focus from informativeness to image-grounded correctness. Experiments on multiple models and benchmarks show that BIRCH reduces informativeness bias by up to 17%, resulting in performance gains of up to 9.8%. Our work reveals an overlooked but fundamental flaw in current VLM-as-a-Judge systems and highlights the need for more principled designs.
Ask Lune
Know this paper in minutes, not an afternoon.
Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.
Lineage
- TruthfulQA: Measuring How Models Mimic Human Falsehoods cites Training language models to follow instructions with human feedback
- Training language models to follow instructions with human feedback cites TruthfulQA: Measuring How Models Mimic Human Falsehoods
- Debating with More Persuasive LLMs Leads to More Truthful Answers cites Learning to summarize with human feedback
Builds on12
- Training language models to follow instructions with human feedbackLong Ouyang, Jeffrey Wu, Xu Jiang, Diogo Almeida et al.NeurIPS 2022 · 24,707 citations
- TruthfulQA: Measuring How Models Mimic Human FalsehoodsStephanie Lin, Jacob Hilton, Owain EvansACL 2022 · 3,228 citations
- Debating with More Persuasive LLMs Leads to More Truthful AnswersAkbir Khan, John Hughes, Dan Valentine, Laura Ruis et al.ICML 2024 · 244 citations
- Learning to summarize with human feedbackNisan Stiennon, Long Ouyang, Jeffrey Wu, Daniel M. Ziegler et al.NeurIPS 2020 · 124 citations
- Benchmarking Cognitive Biases in Large Language Models as EvaluatorsRyan Koo, Minhwa Lee, Vipul Raheja, Jong Inn Park et al.ACL Findings 2024 · 73 citations
Related papers
- MM-JudgeBias: A Benchmark for Evaluating Compositional Biases in MLLM-as-a-JudgeSua Lee, Sanghee Park, Jinbae ImACL 2026 · 1 citation
- Justice or Prejudice? Quantifying Biases in LLM-as-a-JudgeJiayi Ye, Yanbo Wang, Yue Huang, Dongping Chen et al.ICLR 2025
- How Reliable is Multilingual LLM-as-a-Judge?Xiyan Fu, Wei LiuEMNLP Findings 2025 · 7 citations
- PRISM: A Benchmark for Unveiling Cross-modal Knowledge Inconsistency in Large Vision-Language ModelsMingjie Wei, Wei-Nan Zhang, Chen Zhang, Yifeng Ding et al.ACM MM 2025 · 1 citation
- Trust but Verify: Programmatic VLM Evaluation in the WildViraj Prabhu, Senthil Purushwalkam, An Yan, Caiming Xiong et al.ICCV 2025
