mlsanityData to 14 Oct 2026

(How) Do MLLMs Report Bistable Images Like Humans?

arXiv 2609.13254 ↗First posted 6 Sep 2026cs.CV cs.AIcited by 0

Not on the map. Why?

Abstract (arXiv)

Bistable images such as the duck-rabbit are classic stimuli in which one image supports multiple mutually incompatible interpretations, typically reported one at a time in humans. We ask whether multimodal large language models (MLLMs) show similar report behavior and what internal computations support it. Using the LLaVA family, we study two tractable dimensions: modulability, whether reports can be biased by bottom-up visual cues and top-down linguistic priors, and exclusivity, whether responses commit to a single interpretation. We test both on the canonical duck-rabbit and on synthetic Visual Anagrams to mitigate memorization confounds. Behaviorally, both visual and linguistic manipulations systematically shift reports in human-consistent ways, while responses remain predominantly exclusive. Mechanistically, these effects arise from competing image-token representations, distinct pathways for bottom-up and top-down modulation, and a link between exclusive reporting and object-count encoding. Code and data are available at https://github.com/rtakatsky/mllm-bistable-images.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 26 of its 66 references that are in the map. First posted on arXiv 6 Sep 2026.

Builds on

26 of 66 references.

References · 26

  1. 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,744
  2. 2023Improved Baselines with Visual Instruction Tuningcited by 4,418
  3. 2022Large Language Models are Zero-Shot Reasonerscited by 4,374
  4. 2024Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolutioncited by 4,234
  5. 2023InstructBLIP: Towards General-purpose Vision-Language Models with Instruction Tuningcited by 2,876
  6. 2025s1: Simple test-time scalingcited by 1,341
  7. 2023Dissecting Recall of Factual Associations in Auto-Regressive Language Modelscited by 528
  8. 2024Mechanistic Interpretability for AI Safety -- A Reviewcited by 422
  9. 2023Finding Neurons in a Haystack: Case Studies with Sparse Probingcited by 334
  10. 2023What does CLIP know about a red circle? Visual prompt engineering for VLMscited by 218
  11. 2023ViP-LLaVA: Making Large Multimodal Models Understand Arbitrary Visual Promptscited by 161
  12. 2024Devils in Middle Layers of Large Vision-Language Models: Interpreting, Detecting and Mitigating Object Hallucinations via Attention Lenscited by 143
14 more
  1. 2024Interpreting and Editing Vision-Language Representations to Mitigate Hallucinationscited by 104
  2. 2024From Tokens to Words: On the Inner Lexicon of LLMscited by 48
  3. 2023Grounding Visual Illusions in Language: Do Vision-Language Models Perceive Illusions Like Humans?cited by 32
  4. 2025Pixels Versus Priors: Controlling Knowledge Priors in Vision-Language Models through Visual Counterfactscited by 32
  5. 2024Can We Talk Models Into Seeing the World Differently?cited by 23
  6. 2025How Do Vision-Language Models Process Conflicting Information Across Modalities?cited by 23
  7. 2025Hallucinatory Image Tokens: A Training-free EAZY Approach on Detecting and Mitigating Object Hallucinations in LVLMscited by 15
  8. 2024The Illusion-Illusion: Vision Language Models See Illusions Where There Are Nonecited by 14
  9. 2025Understanding Language Prior of LVLMs by Contrasting Chain-of-Embeddingcited by 14
  10. 2024Position: An Inner Interpretability Framework for AI Inspired by Lessons from Cognitive Neurosciencecited by 11
  11. 2024The Art of Deception: Color Visual Illusions and Diffusion Modelscited by 9
  12. 2024Evaluating Vision-Language Models on Bistable Imagescited by 6
  13. 2025Do Large Vision-Language Models Distinguish between the Actual and Apparent Features of Illusions?cited by 3
  14. 2025The Perceptual Observatory Characterizing Robustness and Grounding in MLLMscited by 3

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026Vision-Language Asymmetry in Bistable Image Captioning
  2. 2025Evaluating Graphical Perception with Multimodal LLMs
  3. 2025Where MLLMs Attend and What They Rely On: Explaining Autoregressive Token Generation
  4. 2024The Instinctive Bias: Spurious Images lead to Illusion in MLLMs
  5. 2025MLLMs are Deeply Affected by Modality Bias
  6. 2024A Comprehensive Study of Multimodal Large Language Models for Image Quality Assessment
  7. 2025Mixed Signals: Decoding VLMs' Reasoning and Underlying Bias in Vision-Language Conflict
  8. 2026Seeing or Knowing? Visual Context Sensitivity in Multimodal Large Language Models