mlsanityData to 7 Oct 2026

MM-JudgeBias: A Benchmark for Evaluating Compositional Biases in MLLM-as-a-Judge

arXiv 2604.18164 ↗First posted 20 Apr 2026cs.CL cs.AI cs.CVcited by 3

On the map: Multimodal LLMs / vision-language models → MLLM training and visual instruction tuning

Abstract (arXiv)

Multimodal Large Language Models (MLLMs) have been increasingly used as automatic evaluators-a paradigm known as MLLM-as-a-Judge. However, their reliability and vulnerabilities to biases remain underexplored. We find that many MLLM judges fail to reliably integrate key visual or textual cues, yielding unreliable evaluations when evidence is missing or mismatched, and exhibiting instability under semantically irrelevant perturbations. To address this, we systematically define Compositional Bias in MLLM-as-a-Judge systems and introduce MM-JudgeBias, a benchmark for evaluating it. MM-JudgeBias introduces controlled perturbations across Query, Image, and Response, and evaluates model behavior via two complementary metrics: Bias-Deviation (BD) for sensitivity and Bias-Conformity (BC) for stability. Our dataset of over 1,800 curated and refined multimodal samples, drawn from 29 source benchmarks, enables a fine-grained diagnosis of nine bias types across diverse tasks and domains. Experiments on 26 state-of-the-art MLLMs reveal systematic modality neglect and asymmetric evaluation tendencies, underscoring the need for more reliable judges.

Key facts. Cited by 3 papers in the map (data through 7 Oct 2026). Placed in MLLM training and visual instruction tuning, one of 16 directions in Multimodal LLMs / vision-language models. Builds on 47 of its 62 references that are in the map. First posted on arXiv 20 Apr 2026.

Builds on

47 of 62 references.

Closest ancestors · 11

Same direction, newest first

  1. 2025Multi-Crit: Benchmarking Multimodal Judges on Pluralistic Criteria-Following26 Nov
  2. 2025MM-CRITIC: A Holistic Evaluation of Large Multimodal Models as Multimodal Critique12 Nov
  3. 2025MLLMs are Deeply Affected by Modality Bias24 May
  4. 2025Scaling Text-Rich Image Understanding via Code-Guided Synthetic Multimodal Data Generation20 Feb
  5. 2024Looking Beyond Text: Reducing Language bias in Large Vision-Language Models via Multimodal Dual-Attention and Soft-Image Guidance21 Nov
  6. 2024LLaVA-OneVision: Easy Visual Task Transfer6 Aug
  7. 2024Quantifying and Mitigating Unimodal Biases in Multimodal Large Language Models: A Causal Perspective27 Mar
  8. 2024The Revolution of Multimodal Large Language Models: A Survey19 Feb
  9. 2024MLLM-as-a-Judge: Assessing Multimodal LLM-as-a-Judge with Vision-Language Benchmark7 Feb
  10. 2023Improved Baselines with Visual Instruction Tuning5 Oct
  11. 2023MM-Vet: Evaluating Large Multimodal Models for Integrated Capabilities4 Aug

Other references · 36

  1. 2014Microsoft COCO: Common Objects in Contextcited by 17,498
  2. 2023Judging LLM-as-a-Judge with MT-Bench and Chatbot Arenacited by 8,393
  3. 2022Learn to Explain: Multimodal Reasoning via Thought Chains for Science Question Answeringcited by 1,881
  4. 2019Towards VQA Models That Can Readcited by 1,696
  5. 2023MathVista: Evaluating Mathematical Reasoning of Foundation Models in Visual Contextscited by 1,582
  6. 2025InternVL3.5: Advancing Open-Source Multimodal Models in Versatility, Reasoning, and Efficiencycited by 1,419
  7. 2024A Survey on LLM-as-a-Judgecited by 1,384
  8. 2020DocVQA: A Dataset for VQA on Document Imagescited by 1,288
  9. 2016A Diagram Is Worth A Dozen Imagescited by 793
  10. 2022A-OKVQA: A Benchmark for Visual Question Answering using World Knowledgecited by 702
  11. 2018From Recognition to Cognition: Visual Commonsense Reasoningcited by 553
  12. 2024SeeClick: Harnessing GUI Grounding for Advanced Visual GUI Agentscited by 553
24 more
  1. 2024Measuring Multimodal Mathematical Reasoning with MATH-Vision Datasetcited by 516
  2. 2024MMMU-Pro: A More Robust Multi-discipline Multimodal Understanding Benchmarkcited by 494
  3. 2024LLMs-as-Judges: A Comprehensive Survey on LLM-based Evaluation Methodscited by 451
  4. 2021Inter-GPS: Interpretable Geometry Problem Solving with Formal Language and Symbolic Reasoningcited by 376
  5. 2024Justice or Prejudice? Quantifying Biases in LLM-as-a-Judgecited by 354
  6. 2024Humans or LLMs as the Judge? A Study on Judgement Biasescited by 314
  7. 2024Judging the Judges: A Systematic Study of Position Bias in LLM-as-a-Judgecited by 249
  8. 2022UniGeo: Unifying Geometry Logical Reasoning via Reformulating Mathematical Expressioncited by 137
  9. 2024Prometheus-Vision: Vision-Language Model as a Judge for Fine-Grained Evaluationcited by 123
  10. 2025Preference Leakage: A Contamination Problem in LLM-as-a-judgecited by 119
  11. 2023GPT-4V(ision) as a Generalist Evaluator for Vision-Language Taskscited by 118
  12. 2023ChartBench: A Benchmark for Complex Visual Reasoning in Chartscited by 116
  13. 2025ChartQAPro: A More Diverse and Challenging Benchmark for Chart Question Answeringcited by 105
  14. 2024VL-RewardBench: A Challenging Benchmark for Vision-Language Generative Reward Modelscited by 85
  15. 2022The ArtBench Dataset: Benchmarking Generative Models with Artworkscited by 64
  16. 2025Evaluating Scoring Bias in LLM-as-a-Judgecited by 42
  17. 2024MMCode: Benchmarking Multimodal Large Language Models for Code Generation with Visually Rich Programming Problemscited by 38
  18. 2025Play Favorites: A Statistical Method to Measure Self-Bias in LLM-as-a-Judgecited by 35
  19. 2025Judge Anything: MLLM as a Judge Across Any Modalitycited by 33
  20. 2023A large-scale dataset for end-to-end table recognition in the wildcited by 24
  21. 2025VLRMBench: A Comprehensive and Challenging Benchmark for Vision-Language Reward Modelscited by 19
  22. 2025Judging the Judges: Can Large Vision-Language Models Fairly Evaluate Chart Comprehension and Reasoning?cited by 10
  23. 2025Fooling the LVLM Judges: Visual Biases in LVLM-Based Evaluationcited by 10
  24. 2025Evaluating and Mitigating LLM-as-a-judge Bias in Communication Systemscited by 6

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 3.

3 papers.

  1. 2026REKEY: Metadata-Grounded Visual-Key Regeneration for Contamination-Resilient VQA Evaluationcited by 0
  2. 2026Forged Peer Judgments Mislead Multimodal LLM Judge Panels: Source-Blind Anchoring and Panel-Consensus Verificationcited by 0
  3. 2026Do MLLM Judges Judge the Edit? Auditing Bias in Image Editing Evaluation with Verified Quality Preservationcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026MLLM-as-a-Judge Exhibits Model Preference Bias
  2. 2025Reasoning Multimodal Large Language Model: Data Contamination and Dynamic Evaluation
  3. 2026BabelJudge: Measuring LLM-as-a-Judge Reliability Across Languages and Agent Trajectories
  4. 2026Advancing Multimodal Judge Models through a Capability-Oriented Benchmark and MCTS-Driven Data Generation
  5. 2025MR. Judge: Multimodal Reasoner as a Judge
  6. 2026When Vision-Language Models Judge Without Seeing: Exposing Informativeness Bias
  7. 2023MLLM-Bench: Evaluating Multimodal LLMs with Per-sample Criteria
  8. 2026Mitigating Perceptual Judgment Bias in Multimodal LLM-as-a-Judge via Perceptual Perturbation and Reward Modeling