mlsanityData to 7 Oct 2026

Fooling the LVLM Judges: Visual Biases in LVLM-Based Evaluation

arXiv 2505.15249 ↗First posted 21 May 2025cs.CL cs.CVcited by 10

On the map: Multimodal LLMs / vision-language models → VLM benchmarks and visual reasoning

Abstract (arXiv)

Recently, large vision-language models (LVLMs) have emerged as the preferred tools for judging text-image alignment, yet their robustness along the visual modality remains underexplored. This work is the first study to address a key research question: Can adversarial visual manipulations systematically fool LVLM judges into assigning unfairly inflated scores? We define potential image induced biases within the context of T2I evaluation and examine how these biases affect the evaluations of LVLM judges. Moreover, we introduce a novel, fine-grained, multi-domain meta-evaluation benchmark named FRAME, which is deliberately constructed to exhibit diverse score distributions. By introducing the defined biases into the benchmark, we reveal that all tested LVLM judges exhibit vulnerability across all domains, consistently inflating scores for manipulated images. Further analysis reveals that combining multiple biases amplifies their effects, and pairwise evaluations are similarly susceptible. Moreover, we observe that visual biases persist under prompt-based mitigation strategies, highlighting the vulnerability of current LVLM evaluation systems and underscoring the urgent need for more robust LVLM judges.

Key facts. Cited by 10 papers in the map (data through 7 Oct 2026). Placed in VLM benchmarks and visual reasoning, one of 16 directions in Multimodal LLMs / vision-language models. Builds on 32 of its 42 references that are in the map. First posted on arXiv 21 May 2025.

Builds on

32 of 42 references.

Closest ancestors · 3

Same direction, newest first

  1. 2024Enhancing Visual-Language Modality Alignment in Large Vision Language Models via Self-Improvement24 May
  2. 2024Visual CoT: Advancing Multi-Modal Language Models with a Comprehensive Dataset and Benchmark for Chain-of-Thought Reasoning25 Mar
  3. 2024Prometheus-Vision: Vision-Language Model as a Judge for Fine-Grained Evaluation12 Jan

Other references · 29

  1. 2023Judging LLM-as-a-Judge with MT-Bench and Chatbot Arenacited by 8,393
  2. 2023Improved Baselines with Visual Instruction Tuningcited by 4,362
  3. 2023Lost in the Middle: How Language Models Use Long Contextscited by 3,383
  4. 2016Improved Techniques for Training GANscited by 2,938
  5. 2019Benchmarking Neural Network Robustness to Common Corruptions and Perturbationscited by 2,857
  6. 2024LLaVA-OneVision: Easy Visual Task Transfercited by 2,669
  7. 2021CLIPScore: A Reference-free Evaluation Metric for Image Captioningcited by 2,388
  8. 2023G-Eval: NLG Evaluation using GPT-4 with Better Human Alignmentcited by 2,205
  9. 2024A Survey on LLM-as-a-Judgecited by 1,384
  10. 2023Large Language Models are not Fair Evaluatorscited by 1,050
  11. 2023Human Preference Score v2: A Solid Benchmark for Evaluating Human Preferences of Text-to-Image Synthesiscited by 847
  12. 2024Evaluating Text-to-Visual Generation with Image-to-Text Generationcited by 465
17 more
  1. 2023TIFA: Accurate and Interpretable Text-to-Image Faithfulness Evaluation with Question Answeringcited by 416
  2. 2024MLLM-as-a-Judge: Assessing Multimodal LLM-as-a-Judge with Vision-Language Benchmarkcited by 362
  3. 2023JudgeLM: Fine-tuned Large Language Models are Scalable Judgescited by 351
  4. 2023Set-of-Mark Prompting Unleashes Extraordinary Visual Grounding in GPT-4Vcited by 291
  5. 2023Evaluating Large Language Models at Evaluating Instruction Followingcited by 290
  6. 2023What does CLIP know about a red circle? Visual prompt engineering for VLMscited by 215
  7. 2023VIEScore: Towards Explainable Metrics for Conditional Image Synthesis Evaluationcited by 196
  8. 2023GPT-4V(ision) as a Generalist Evaluator for Vision-Language Taskscited by 118
  9. 2024Is LLM-as-a-Judge Robust? Investigating Universal Adversarial Attacks on Zero-shot LLM Assessmentcited by 117
  10. 2023What You See is What You Read? Improving Text-Image Alignment Evaluationcited by 105
  11. 2024ConceptMix: A Compositional Image Generation Benchmark with Controllable Difficultycited by 60
  12. 2020Adv-watermark: A Novel Watermark Perturbation for Adversarial Examplescited by 56
  13. 2024Visual Agents as Fast and Slow Thinkerscited by 53
  14. 2024Are LLM-Judges Robust to Expressions of Uncertainty? Investigating the effect of Epistemic Markers on LLM-based Evaluationcited by 29
  15. 2025Benchmarking Adversarial Robustness to Bias Elicitation in Large Language Models: Scalable Automated Assessment with LLM-as-a-Judgecited by 23
  16. 2024Scaling Trends in Language Model Robustnesscited by 18
  17. 2025LLMs can be easily Confused by Instructional Distractionscited by 12

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2025: 1; 2026: 9.

10 papers.

  1. 2025Presenting a Paper is an Art: Self-Improvement Aesthetic Agents for Academic Presentationscited by 11
  2. 2026MM-JudgeBias: A Benchmark for Evaluating Compositional Biases in MLLM-as-a-Judgecited by 3
  3. 2026On the Adversarial Robustness of Multimodal LLM Judgescited by 3
  4. 2026Prototypicality Bias Reveals Blindspots in Multimodal Evaluation Metricscited by 2
  5. 2026Do Image-Text Metrics Respect Semantic Invariances?cited by 1
  6. 2026Forged Peer Judgments Mislead Multimodal LLM Judge Panels: Source-Blind Anchoring and Panel-Consensus Verificationcited by 0
  7. 2026JoLT: Joint Latent Trajectories for Context-Guided High-Resolution Tiled Generationcited by 0
  8. 2026A Low Grounding Score Is Not an Ungrounded Judge: Identifying the Perceptibility Confound in Multimodal Oversightcited by 0
  9. 2026When the Judge Acts: Auditing VLM-Guided Image Selection on Culturally Situated Promptscited by 0
  10. 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. 2026Language Bias in LVLMs: From In-Depth Analysis to Simple and Effective Mitigation
  2. 2024The Instinctive Bias: Spurious Images lead to Illusion in MLLMs
  3. 2025Mixed Signals: Decoding VLMs' Reasoning and Underlying Bias in Vision-Language Conflict
  4. 2025Improving Alignment in LVLMs with Debiased Self-Judgment
  5. 2025MVI-Bench: A Comprehensive Benchmark for Evaluating Robustness to Misleading Visual Inputs in LVLMs
  6. 2026V-FAT: Benchmarking Visual Fidelity Against Text-bias
  7. 2024Uncovering Bias in Large Vision-Language Models at Scale with Counterfactuals
  8. 2026Seeing Isn't Believing: Uncovering Blind Spots in Evaluator Vision-Language Models