mlsanityData to 14 Oct 2026

MM-SpuBench: Towards Better Understanding of Spurious Biases in Multimodal LLMs

arXiv 2406.17126 ↗First posted 24 Jun 2024cs.CV cs.LGcited by 33

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

Abstract (arXiv)

Spurious bias, a tendency to exploit spurious correlations between superficial input attributes and prediction targets, has revealed a severe robustness pitfall in classical machine learning problems. Multimodal Large Language Models (MLLMs), which leverage pretrained vision and language models, have recently demonstrated strong capability in joint vision-language understanding. However, both the presence and severity of spurious biases in MLLMs remain poorly understood. In this work, we address this gap by analyzing the spurious biases in the multimodal setting and uncovering the specific inference-time data patterns that can manifest this problem. To support this analysis, we introduce MM-SpuBench, a comprehensive, human-verified benchmark dataset consisting of image-class pairs annotated with core and spurious attributes, grounded in our taxonomy of nine distinct types of spurious correlations. The benchmark is constructed using human-interpretable attribute information to capture a wide range of spurious patterns reflective of real-world knowledge. Leveraging this benchmark, we conduct a comprehensive evaluation of the state-of-the-art open-source and proprietary MLLMs with both standard accuracy and the proposed Conditional Generation Likelihood Advantage (CGLA). Our findings highlight the persistence of reliance on spurious correlations and the difficulty of mitigation on our benchmark. We hope this work can inspire new technical strides to mitigate these biases. Our benchmark is publicly available at https://huggingface.co/datasets/mmbench/MM-SpuBench.

Key facts. Cited by 33 papers in the map (data through 14 Oct 2026). Placed in MLLM training and visual instruction tuning, one of 16 directions in Multimodal LLMs / vision-language models. Builds on 63 of its 92 references that are in the map. First posted on arXiv 24 Jun 2024.

Builds on

63 of 92 references.

Closest ancestors · 8

Same direction, newest first

  1. 2024MM1: Methods, Analysis & Insights from Multimodal LLM Pre-training14 Mar
  2. 2024MLLM-as-a-Judge: Assessing Multimodal LLM-as-a-Judge with Vision-Language Benchmark7 Feb
  3. 2024Eyes Wide Shut? Exploring the Visual Shortcomings of Multimodal LLMs11 Jan
  4. 2023InternVL: Scaling up Vision Foundation Models and Aligning for Generic Visual-Linguistic Tasks21 Dec
  5. 2023VisIT-Bench: A Benchmark for Vision-Language Instruction Following Inspired by Real-World Use12 Aug
  6. 2023MM-Vet: Evaluating Large Multimodal Models for Integrated Capabilities4 Aug
  7. 2023MMBench: Is Your Multi-modal Model an All-around Player?12 Jul
  8. 2023InstructBLIP: Towards General-purpose Vision-Language Models with Instruction Tuning11 May

Other references · 55

  1. 2018BERT: Pre-training of Deep Bidirectional Transformers for Language Understandingcited by 40,399
  2. 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,744
  3. 2023GPT-4 Technical Reportcited by 18,396
  4. 2023LLaMA: Open and Efficient Foundation Language Modelscited by 14,132
  5. 2023Llama 2: Open Foundation and Fine-Tuned Chat Modelscited by 13,028
  6. 2023Judging LLM-as-a-Judge with MT-Bench and Chatbot Arenacited by 8,626
  7. 2023DINOv2: Learning Robust Visual Features without Supervisioncited by 7,974
  8. 2023Visual Instruction Tuningcited by 7,869
  9. 2021Masked Autoencoders Are Scalable Vision Learnerscited by 7,403
  10. 2016Grad-CAM: Visual Explanations from Deep Networks via Gradient-based Localizationcited by 6,733
  11. 2022PaLM: Scaling Language Modeling with Pathwayscited by 5,326
  12. 2021Finetuned Language Models Are Zero-Shot Learnerscited by 3,780
38 more
  1. 2022OPT: Open Pre-trained Transformer Language Modelscited by 3,512
  2. 2023Qwen Technical Reportcited by 3,344
  3. 2022Scaling Instruction-Finetuned Language Modelscited by 3,024
  4. 2019Benchmarking Neural Network Robustness to Common Corruptions and Perturbationscited by 2,877
  5. 2020Shortcut Learning in Deep Neural Networkscited by 1,969
  6. 2020The Many Faces of Robustness: A Critical Analysis of Out-of-Distribution Generalizationcited by 1,850
  7. 2023Evaluating Object Hallucination in Large Vision-Language Modelscited by 1,729
  8. 2019Towards VQA Models That Can Readcited by 1,718
  9. 2018ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustnesscited by 1,432
  10. 2019Natural Adversarial Examplescited by 1,408
  11. 2019Learning Robust Global Representations by Penalizing Local Predictive Powercited by 1,013
  12. 2024Yi: Open Foundation Models by 01.AIcited by 719
  13. 2023CogVLM: Visual Expert for Pretrained Language Modelscited by 708
  14. 2023Language Is Not All You Need: Aligning Perception with Language Modelscited by 543
  15. 2021Just Train Twice: Improving Group Robustness without Training Group Informationcited by 523
  16. 2023V*: Guided Visual Search as a Core Mechanism in Multimodal LLMscited by 482
  17. 2023OPERA: Alleviating Hallucination in Multi-Modal Large Language Models via Over-Trust Penalty and Retrospection-Allocationcited by 478
  18. 2023GLaMM: Pixel Grounding Large Multimodal Modelcited by 455
  19. 2024InternLM2 Technical Reportcited by 397
  20. 2022Last Layer Re-Training is Sufficient for Robustness to Spurious Correlationscited by 381
  21. 2020Noise or Signal: The Role of Image Backgrounds in Object Recognitioncited by 343
  22. 2020An Investigation of Why Overparameterization Exacerbates Spurious Correlationscited by 310
  23. 2019Don't Take the Easy Way Out: Ensemble Based Methods for Avoiding Known Dataset Biasescited by 280
  24. 2023LAMM: Language-Assisted Multi-Modal Instruction-Tuning Dataset, Framework, and Benchmarkcited by 169
  25. 2023Change is Hard: A Closer Look at Subpopulation Shiftcited by 120
  26. 2019SemanticAdv: Generating Adversarial Examples via Attribute-conditional Image Editingcited by 93
  27. 2023On the Robustness of Large Multimodal Models Against Image Adversarial Attackscited by 92
  28. 2022Spread Spurious Attribute: Improving Worst-group Accuracy with Spurious Attribute Estimationcited by 88
  29. 2021Identifying and Mitigating Spurious Correlations for Improving Robustness in NLP Modelscited by 76
  30. 2023Discover and Cure: Concept-aware Mitigation of Spurious Correlationcited by 71
  31. 2021On the Impact of Spurious Correlation for Out-of-distribution Detectioncited by 68
  32. 2023Seeing is not always believing: Benchmarking Human and Model Perception of AI-Generated Imagescited by 66
  33. 2021Meaningfully Debugging Model Mistakes using Conceptual Counterfactual Explanationscited by 53
  34. 2018Manifold: A Model-Agnostic Framework for Interpretation and Diagnosis of Machine Learning Modelscited by 50
  35. 2018Towards Accountable AI: Hybrid Human-Machine Analyses for Characterizing System Failurecited by 49
  36. 2023Mitigating Spurious Correlations in Multi-modal Models during Fine-tuningcited by 47
  37. 2023LANCE: Stress-testing Visual Models by Generating Language-guided Counterfactual Imagescited by 46
  38. 2021Finding and Fixing Spurious Patterns with Explanationscited by 38

50 most cited of 55 All 55 (JSON)

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2024: 8; 2025: 17; 2026: 8.

33 papers.

Sort:

By area:

  1. 2024A Survey on Benchmarks of Multimodal Large Language Modelscited by 82
  2. 2024MME-Survey: A Comprehensive Survey on Evaluation of Multimodal LLMscited by 72
  3. 2024Visual Prompting in Multimodal Large Language Models: A Surveycited by 54
  4. 2024A Survey on Evaluation of Multimodal Large Language Modelscited by 46
  5. 2024A Survey on Multimodal Benchmarks: In the Era of Large AI Modelscited by 27
  6. 2024Unified Parameter-Efficient Unlearning for LLMscited by 25
  7. 2025Do Vision-Language Models Have Internal World Models? Towards an Atomic Evaluationcited by 24
  8. 2025HumaniBench: A Human-Centric Framework for Large Multimodal Models Evaluationcited by 19
  9. 2025Safety Mirage: How Spurious Correlations Undermine VLM Safety Fine-Tuning and Can Be Mitigated by Machine Unlearningcited by 16
  10. 2025Escaping the SpuriVerse: Can Large Vision-Language Models Generalize Beyond Seen Spurious Correlations?cited by 14
  11. 2025Insect-Foundation: A Foundation Model and Large Multimodal Dataset for Vision-Language Insect Understandingcited by 7
  12. 2025SpurLens: Automatic Detection of Spurious Cues in Multimodal LLMscited by 7
21 more
  1. 2025Structured Spectral Reasoning for Frequency-Adaptive Multimodal Recommendationcited by 6
  2. 2025Visual Instruction Bottleneck Tuningcited by 5
  3. 2025Directed-Tokens: A Robust Multi-Modality Alignment Approach to Large Language-Vision Modelscited by 5
  4. 2025Beyond Spurious Signals: Debiasing Multimodal Large Language Models via Counterfactual Inference and Adaptive Expert Routingcited by 5
  5. 2024Measuring Agreeableness Bias in Multimodal Modelscited by 4
  6. 2025GHOST: Hallucination-Inducing Image Generation for Multimodal LLMscited by 4
  7. 2025SNS-Bench-VL: Benchmarking Multimodal Large Language Models in Social Networking Servicescited by 2
  8. 2025Bias Beyond Demographics: Probing Decision Boundaries in Black-Box LVLMs via Counterfactual VQAcited by 2
  9. 2025Seeing Beyond the Scene: Analyzing and Mitigating Background Bias in Action Recognitioncited by 2
  10. 2025Unbiased Visual Reasoning with Controlled Visual Inputscited by 2
  11. 2026Prototypicality Bias Reveals Blindspots in Multimodal Evaluation Metricscited by 2
  12. 2025The Role of Model Confidence on Bias Effects in Measured Uncertainties for Vision-Language Modelscited by 1
  13. 2025SAGE: Spuriousness-Aware Guided Prompt Exploration for Mitigating Multimodal Biascited by 1
  14. 2026FiLoRA: Focus-and-Ignore LoRA for Controllable Feature Reliancecited by 1
  15. 2024Verb Mirage: Unveiling and Assessing Verb Concept Hallucinations in Multimodal Large Language Modelscited by 0
  16. 2026Evaluating Stochastic Collapse and Implicit Bias in Multimodal Large Language Modelscited by 0
  17. 2026MotionHalluc: Diagnosing Kinematic Hallucinations in Fine-Grained Motion Reasoningcited by 0
  18. 2026MissingBench-Verified: Probing Vision-Language Models' Inability to Detect Missing Object Partscited by 0
  19. 2026Partition-Aware Unlearning for Removing Spurious Correlations in Large Vision-Language Modelscited by 0
  20. 2026GUIDE: Guiding Internal Evidence with Language Instructionscited by 0
  21. 2026Visual Input and Its Framing Affect Attribute-based Descriptions Produced by Large Vision-Language Modelscited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2023MultiModal Bias: Introducing a Framework for Stereotypical Bias Assessment beyond Gender and Race in Vision Language Models
  2. 2024The Instinctive Bias: Spurious Images lead to Illusion in MLLMs
  3. 2021Worst of Both Worlds: Biases Compound in Pre-trained Vision-and-Language Models
  4. 2024M5 -- A Diverse Benchmark to Assess the Performance of Large Multimodal Models Across Multilingual and Multicultural Vision-Language Tasks
  5. 2026Multimodal Large Language Models as Image Classifiers
  6. 2026MM-JudgeBias: A Benchmark for Evaluating Compositional Biases in MLLM-as-a-Judge
  7. 2024VLBiasBench: A Comprehensive Benchmark for Evaluating Bias in Large Vision-Language Model
  8. 2024Uncovering Bias in Large Vision-Language Models at Scale with Counterfactuals