mlsanityData to 7 Oct 2026

Mechanisms of Prompt-Induced Hallucination in Vision-Language Models

arXiv 2601.05201 ↗First posted 8 Jan 2026cs.CV cs.AI cs.CLcited by 4

On the map: Multimodal LLMs / vision-language models → Hallucination mitigation in LVLMs

Abstract (arXiv)

Large vision-language models (VLMs) are highly capable, yet often hallucinate by favoring textual prompts over visual evidence. We study this failure mode in a controlled object-counting setting, where the prompt overstates the number of objects in the image (e.g., asking a model to describe four waterlilies when only three are present). At low object counts, models often correct the overestimation, but as the number of objects increases, they increasingly conform to the prompt regardless of the discrepancy. Through mechanistic analysis of three VLMs, we identify a small set of attention heads whose ablation substantially reduces prompt-induced hallucinations (PIH) by at least 40% without additional training. Across models, PIH-heads mediate prompt copying in model-specific ways. We characterize these differences and show that PIH ablation increases correction toward visual evidence. Our findings offer insights into the internal mechanisms driving prompt-induced hallucinations, revealing model-specific differences in how these behaviors are implemented.

Key facts. Cited by 4 papers in the map (data through 7 Oct 2026). Placed in Hallucination mitigation in LVLMs, one of 16 directions in Multimodal LLMs / vision-language models. Builds on 40 of its 54 references that are in the map. First posted on arXiv 8 Jan 2026.

Builds on

40 of 54 references.

Closest ancestors · 1

Same direction, newest first

  1. 2025What Color Is the Text? A Benchmark for Hallucination Induced by Image-Embedded Prompt17 Nov

Other references · 39

  1. 2019PyTorch: An Imperative Style, High-Performance Deep Learning Librarycited by 20,160
  2. 2024Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolutioncited by 4,151
  3. 2024LLaVA-OneVision: Easy Visual Task Transfercited by 2,669
  4. 2022Locating and Editing Factual Associations in GPTcited by 2,609
  5. 2019HuggingFace's Transformers: State-of-the-art Natural Language Processingcited by 2,201
  6. 2023MM-Vet: Evaluating Large Multimodal Models for Integrated Capabilitiescited by 1,221
  7. 2022Interpretability in the Wild: a Circuit for Indirect Object Identification in GPT-2 smallcited by 1,130
  8. 2019Attention is not Explanationcited by 916
  9. 2025Janus-Pro: Unified Multimodal Understanding and Generation with Data and Model Scalingcited by 787
  10. 2019Is Attention Interpretable?cited by 387
  11. 2023Adaptive Chameleon or Stubborn Sloth: Revealing the Behavior of Large Language Models in Knowledge Conflictscited by 333
  12. 2023How Language Model Hallucinations Can Snowballcited by 310
27 more
  1. 2024Knowledge Conflicts for LLMs: A Surveycited by 274
  2. 2023Teaching CLIP to Count to Tencited by 194
  3. 2023LLM Lies: Hallucinations are not Bugs, but Features as Adversarial Examplescited by 177
  4. 2024Fine-Tuning Enhances Existing Mechanisms: A Case Study on Entity Trackingcited by 140
  5. 2019On Identifiability in Transformerscited by 115
  6. 2023The Earth is Flat because...: Investigating LLMs' Belief towards Misinformation via Persuasive Conversationcited by 107
  7. 2024Cognitive Bias in Decision-Making with LLMscited by 104
  8. 2022Capturing Failures of Large Language Models via Human Cognitive Biasescited by 97
  9. 2025Words or Vision: Do Vision-Language Models Have Blind Faith in Text?cited by 86
  10. 2021Vision-and-Language or Vision-for-Language? On Cross-Modal Influence in Multimodal Transformerscited by 71
  11. 2025Vision Language Models are Biasedcited by 63
  12. 2024Measuring and Controlling Instruction (In)Stability in Language Model Dialogscited by 51
  13. 2025MIB: A Mechanistic Interpretability Benchmarkcited by 49
  14. 2023Personas as a Way to Model Truthfulness in Language Modelscited by 46
  15. 2024IsoBench: Benchmarking Multimodal Foundation Models on Isomorphic Representationscited by 37
  16. 2025Language Models use Lookbacks to Track Beliefscited by 36
  17. 2025Same Task, Different Circuits: Disentangling Modality-Specific Mechanisms in VLMscited by 35
  18. 2023ReFACT: Updating Text-to-Image Models by Editing the Text Encodercited by 32
  19. 2024A Comprehensive Evaluation of Cognitive Biases in LLMscited by 32
  20. 2025Pixels Versus Priors: Controlling Knowledge Priors in Vision-Language Models through Visual Counterfactscited by 30
  21. 2025Multimodal Inconsistency Reasoning (MMIR): A New Benchmark for Multimodal Reasoning Modelscited by 24
  22. 2025The Dual-Route Model of Inductioncited by 22
  23. 2025How Do Vision-Language Models Process Conflicting Information Across Modalities?cited by 22
  24. 2025When Modalities Conflict: How Unimodal Reasoning Uncertainty Governs Preference Dynamics in MLLMscited by 18
  25. 2025Forgotten Polygons: Multimodal Large Language Models are Shape-Blindcited by 16
  26. 2025Same Content, Different Answers: Cross-Modal Inconsistency in MLLMscited by 7
  27. 2025HACK: Hallucinations Along Certainty and Knowledge Axescited by 6

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 4.

4 papers.

  1. 2026Dual-Pathway Circuits of Object Hallucination in Vision-Language Modelscited by 4
  2. 2026Geometric Risk Control for Vision-Language Model OCRcited by 1
  3. 2026Anchored, Not Graded: Vision-Language Models Fail at Slant-from-Texture Perceptioncited by 0
  4. 2026Prior Directions: Why GUI Grounding Gets Locked in the Pastcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2024Devils in Middle Layers of Large Vision-Language Models: Interpreting, Detecting and Mitigating Object Hallucinations via Attention Lens
  2. 2026When Prompts Override Vision: Prompt-Induced Hallucinations in LVLMs
  3. 2024Multi-Object Hallucination in Vision-Language Models
  4. 2026Reducing Object Hallucination in LVLMs via Emphasizing Image-negative Tokens
  5. 2026Focus Matters: Phase-Aware Suppression for Hallucination in Vision-Language Models
  6. 2025The Role of Background Information in Reducing Object Hallucination in Vision-Language Models: Insights from Cutoff API Prompting
  7. 2026Risk-aware Selective Prompting for Hallucination Mitigation in Large Vision-Language Models
  8. 2025Two Causes, Not One: Rethinking Omission and Fabrication Hallucinations in MLLMs