mlsanityData to 14 Oct 2026

Measuring Epistemic Humility in Multimodal Large Language Models

arXiv 2509.09658 ↗First posted 11 Sep 2025cs.CVcited by 4

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

Abstract (arXiv)

Hallucinations in multimodal large language models (MLLMs), where a model generates content inconsistent with the input image, pose significant risks in real-world applications. Existing hallucination benchmarks primarily test recognition, i.e., whether a model can select the correct answer among distractors. They overlook another capability that is essential for trustworthy AI: recognizing when none of the provided options is supported by the image and declining to commit to a false choice. We present HumbleBench, a hallucination benchmark that evaluates this false-option rejection behavior, a directly observable facet of epistemic humility, through multiple-choice questions that always include a "None of the above" (NOTA) option. HumbleBench is built from the Panoptic Scene Graph dataset: its object and relation annotations, together with candidate attribute cues, are used to prompt GPT-4-Turbo to generate questions, which are then manually verified, yielding 22,831 questions on object, relation, and attribute hallucinations. We evaluate 27 MLLMs, including general-purpose, reasoning-oriented, and proprietary models, and complement standard accuracy with NOTA-aware metrics, diagnostic stress tests, and a human evaluation. Even the strongest models select NOTA for only about half of the questions in which it is correct, cautious prompting shifts models toward NOTA without improving their ability to detect unsupported options, and humans reject unsupported options considerably more reliably than all evaluated models. These results show that recognition accuracy alone is insufficient for assessing hallucination robustness. Our code and dataset are publicly available at https://github.com/maifoundations/HumbleBench.

Key facts. Cited by 4 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 43 of its 55 references that are in the map. First posted on arXiv 11 Sep 2025.

Builds on

43 of 55 references.

Closest ancestors · 11

Same direction, newest first

  1. 2024DeepSeek-VL2: Mixture-of-Experts Vision-Language Models for Advanced Multimodal Understanding13 Dec
  2. 2024Pixtral 12B9 Oct
  3. 2024MME-RealWorld: Could Your Multimodal LLM Challenge High-Resolution Real-World Scenarios that are Difficult for Humans?23 Aug
  4. 2024Cambrian-1: A Fully Open, Vision-Centric Exploration of Multimodal LLMs24 Jun
  5. 2024Ovis: Structural Embedding Alignment for Multimodal Large Language Model31 May
  6. 2023InternVL: Scaling up Vision Foundation Models and Aligning for Generic Visual-Linguistic Tasks21 Dec
  7. 2023VCoder: Versatile Vision Encoders for Multimodal Large Language Models21 Dec
  8. 2023VILA: On Pre-training for Visual Language Models12 Dec
  9. 2023Monkey: Image Resolution and Text Label Are Important Things for Large Multi-modal Models11 Nov
  10. 2023MME: A Comprehensive Evaluation Benchmark for Multimodal Large Language Models23 Jun
  11. 2023InstructBLIP: Towards General-purpose Vision-Language Models with Instruction Tuning11 May

Other references · 32

  1. 2024The Llama 3 Herd of Modelscited by 14,715
  2. 2020Measuring Massive Multitask Language Understandingcited by 7,963
  3. 2025Qwen2.5-VL Technical Reportcited by 5,587
  4. 2022Survey of Hallucination in Natural Language Generationcited by 2,075
  5. 2024Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameterscited by 1,986
  6. 2025InternVL3: Exploring Advanced Training and Test-Time Recipes for Open-Source Multimodal Modelscited by 1,834
  7. 2025Gemma 3 Technical Reportcited by 1,824
  8. 2023Evaluating Object Hallucination in Large Vision-Language Modelscited by 1,729
  9. 2024Phi-4 Technical Reportcited by 711
  10. 2023Mitigating Object Hallucinations in Large Vision-Language Models through Visual Contrastive Decodingcited by 688
  11. 2023HallusionBench: An Advanced Diagnostic Suite for Entangled Language Hallucination and Visual Illusion in Large Vision-Language Modelscited by 583
  12. 2023OPERA: Alleviating Hallucination in Multi-Modal Large Language Models via Over-Trust Penalty and Retrospection-Allocationcited by 478
20 more
  1. 2024LLaVA-CoT: Let Vision Language Models Reason Step-by-Stepcited by 454
  2. 2023Mitigating Hallucination in Large Multi-Modal Models via Robust Instruction Tuningcited by 445
  3. 2025R1-Onevision: Advancing Generalized Multimodal Reasoning through Cross-Modal Formalizationcited by 370
  4. 2024A Survey on Hallucination in Large Vision-Language Modelscited by 357
  5. 2024Molmo and PixMo: Open Weights and Open Data for State-of-the-Art Vision-Language Modelscited by 329
  6. 2025R1-VL: Learning to Reason with Multimodal Large Language Models via Step-wise Group Relative Policy Optimizationcited by 301
  7. 2024Multi-Modal Hallucination Control by Visual Information Groundingcited by 208
  8. 2024Mitigating Hallucinations in Large Vision-Language Models with Instruction Contrastive Decodingcited by 203
  9. 2024Mulberry: Empowering MLLM with o1-like Reasoning and Reflection via Collective Monte Carlo Tree Searchcited by 175
  10. 2024Building and better understanding vision-language models: insights and future directionscited by 173
  11. 2023Hallucination Augmented Contrastive Learning for Multimodal Large Language Modelcited by 121
  12. 2022Panoptic Scene Graph Generationcited by 114
  13. 2025Visionary-R1: Mitigating Shortcuts in Visual Reasoning with Reinforcement Learningcited by 87
  14. 2024Can AI Assistants Know What They Don't Know?cited by 50
  15. 2024THRONE: An Object-based Hallucination Benchmark for the Free-form Generations of Large Vision-Language Modelscited by 36
  16. 2024HaloQuest: A Visual Hallucination Dataset for Advancing Multimodal Reasoningcited by 32
  17. 2024Reefknot: A Comprehensive Benchmark for Relation Hallucination Evaluation, Analysis and Mitigation in Multimodal Large Language Modelscited by 30
  18. 2024Hallu-PI: Evaluating Hallucination in Multi-modal Large Language Models within Perturbed Inputscited by 14
  19. 2025DASH: Detection and Assessment of Systematic Hallucinations of VLMscited by 14
  20. 2024LongHalQA: Long-Context Hallucination Evaluation for MultiModal Large Language Modelscited by 7

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 2; 2026: 2.

4 papers.

  1. 2025Honesty over Accuracy: Trustworthy Language Models through Reinforced Hesitationcited by 5
  2. 2026Positive Alignment: Artificial Intelligence for Human Flourishingcited by 5
  3. 2025Learning to Think Fast and Slow for Visual Language Modelscited by 3
  4. 2026Accurate but Not Humble: Evaluating Epistemic Humility in LLM Agents under Knowledge Conflictcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2025MoHoBench: Assessing Honesty of Multimodal Large Language Models via Unanswerable Visual Questions
  2. 2024MMEvalPro: Calibrating Multimodal Benchmarks Towards Trustworthy and Efficient Evaluation
  3. 2024TUBench: Benchmarking Large Vision-Language Models on Trustworthiness with Unanswerable Questions
  4. 2026FINER: MLLMs Hallucinate under Fine-grained Negative Queries
  5. 2026Benchmarking Deflection and Hallucination in Large Vision-Language Models
  6. 2026BEAR-Bench: A Bilingual Enterprise and Academic Reasoning Benchmark for Multimodal Models
  7. 2025Taming Object Hallucinations with Verified Atomic Confidence Estimation
  8. 2026SAVOR: Self-Aware Visual Grounding via Confidence-Calibrated Reinforcement Learning for Multimodal Hallucination Mitigation