mlsanityData to 14 Oct 2026

Machine Mental Imagery: Empower Multimodal Reasoning with Latent Visual Tokens

arXiv 2506.17218 ↗First posted 20 Jun 2025cs.CV cs.AIcited by 124

On the map: LLM reasoning → Multimodal reasoning

also in 1 more
Abstract (arXiv)

Vision-language models (VLMs) excel at multimodal understanding, yet their text-only decoding forces them to verbalize visual reasoning, limiting performance on tasks that demand visual imagination. Recent attempts train VLMs to render explicit images, but the heavy image-generation pre-training often hinders the reasoning ability. Inspired by the way humans reason with mental imagery-the internal construction and manipulation of visual cues-we investigate whether VLMs can reason through interleaved multimodal trajectories without producing explicit images. To this end, we present a Machine Mental Imagery framework, dubbed as Mirage, which augments VLM decoding with latent visual tokens alongside ordinary text. Concretely, whenever the model chooses to ``think visually'', it recasts its hidden states as next tokens, thereby continuing a multimodal trajectory without generating pixel-level images. Begin by supervising the latent tokens through distillation from ground-truth image embeddings, we then switch to text-only supervision to make the latent trajectory align tightly with the task objective. A subsequent reinforcement learning stage further enhances the multimodal reasoning capability. Experiments on diverse benchmarks demonstrate that Mirage unlocks stronger multimodal reasoning without explicit image generation.

Key facts. Cited by 124 papers in the map (data through 14 Oct 2026). Placed in Multimodal reasoning, one of 17 directions in LLM reasoning. Builds on 40 of its 44 references that are in the map. First posted on arXiv 20 Jun 2025.

Builds on

40 of 44 references.

Closest ancestors · 12

Same direction, newest first

  1. 2025Thinking with Generated Images28 May
  2. 2025Visual Thoughts: A Unified Perspective of Understanding Multimodal Chain-of-Thought21 May
  3. 2025OpenThinkIMG: Learning to Think with Images via Visual Tool Reinforcement Learning13 May
  4. 2025Think or Not Think: A Study of Explicit Thinking in Rule-Based Visual Reinforcement Fine-Tuning20 Mar
  5. 2025LMM-R1: Empowering 3B LMMs with Strong Reasoning Abilities Through Two-Stage Rule-Based RL10 Mar
  6. 2025Imagine while Reasoning in Space: Multimodal Visualization-of-Thought13 Jan
  7. 2024CoMT: A Novel Benchmark for Chain of Multi-modal Thought on Large Vision-Language Models17 Dec
  8. 2024Visual Sketchpad: Sketching as a Visual Chain of Thought for Multimodal Language Models13 Jun
  9. 2024Image-of-Thought Prompting for Visual Reasoning Refinement in Multimodal Large Language Models22 May
  10. 2024Cantor: Inspiring Multimodal Chain-of-Thought of MLLM24 Apr
  11. 2024Visual CoT: Advancing Multi-Modal Language Models with a Comprehensive Dataset and Benchmark for Chain-of-Thought Reasoning25 Mar
  12. 2023Multimodal Chain-of-Thought Reasoning in Language Models2 Feb

Other references · 28

  1. 2022Chain-of-Thought Prompting Elicits Reasoning in Large Language Modelscited by 12,459
  2. 2024DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Modelscited by 9,379
  3. 2025Qwen2.5-VL Technical Reportcited by 5,587
  4. 2024Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolutioncited by 4,234
  5. 2024CogVideoX: Text-to-Video Diffusion Models with An Expert Transformercited by 2,261
  6. 2024HybridFlow: A Flexible and Efficient RLHF Frameworkcited by 2,057
  7. 2024Chameleon: Mixed-Modal Early-Fusion Foundation Modelscited by 927
  8. 2024Training Large Language Models to Reason in a Continuous Latent Spacecited by 797
  9. 2024BLINK: Multimodal Large Language Models Can See but Not Perceivecited by 550
  10. 2025BLIP3-o: A Family of Fully Open Unified Multimodal Models-Architecture, Training and Datasetcited by 376
  11. 2023Towards Revealing the Mystery behind Chain of Thought: A Theoretical Perspectivecited by 294
  12. 2023Think before you speak: Training Language Models With Pause Tokenscited by 264
16 more
  1. 2024From Explicit CoT to Implicit CoT: Learning to Internalize CoT Step by Stepcited by 235
  2. 2024MetaMorph: Multimodal Understanding and Generation via Instruction Tuningcited by 206
  3. 2023Implicit Chain of Thought Reasoning via Knowledge Distillationcited by 176
  4. 2024Do Large Language Models Latently Perform Multi-Hop Reasoning?cited by 169
  5. 2024Distilling System 2 into System 1cited by 159
  6. 2025Reinforcing Spatial Reasoning in Vision-Language Models with Interwoven Thinking and Visual Drawingcited by 136
  7. 2024ANOLE: An Open, Autoregressive, Native Large Multimodal Models for Interleaved Image-Text Generationcited by 114
  8. 2024Chain of Preference Optimization: Improving Chain-of-Thought Reasoning in LLMscited by 96
  9. 2024Hopping Too Late: Exploring the Limitations of Large Language Models on Multi-Hop Queriescited by 78
  10. 20243D-Mem: 3D Scene Memory for Embodied Exploration and Reasoningcited by 75
  11. 2023Guiding Language Model Reasoning with Planning Tokenscited by 58
  12. 2025Satori: Reinforcement Learning with Chain-of-Action-Thought Enhances LLM Reasoning via Autoregressive Searchcited by 53
  13. 2024VSP: Assessing the dual challenges of perception and reasoning in spatial planning tasks for VLMscited by 47
  14. 2025Teaching Language Models to Critique via Reinforcement Learningcited by 42
  15. 2025From Intention to Execution: Probing the Generalization Boundaries of Vision-Language-Action Modelscited by 23
  16. 2025Autoregressive Semantic Visual Reconstruction Helps VLMs Understand Bettercited by 17

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 28; 2026: 96.

124 papers.

Sort:

By area:

  1. 2025Latent Visual Reasoningcited by 104
  2. 2025Monet: Reasoning in Latent Visual Space Beyond Images and Languagecited by 82
  3. 2025Chain-of-Visual-Thought: Teaching VLMs to See and Think Better with Continuous Visual Tokenscited by 71
  4. 2025Think with 3D: Geometric Imagination Grounded Spatial Reasoning from Limited Viewscited by 65
  5. 2025ThinkMorph: Emergent Properties in Multimodal Interleaved Chain-of-Thought Reasoningcited by 56
  6. 2026Zooming without Zooming: Region-to-Image Distillation for Fine-Grained Multimodal Perceptioncited by 50
  7. 2025Latent Sketchpad: Sketching Visual Thoughts to Elicit Multimodal Reasoning in MLLMscited by 43
  8. 2026The Latent Space: Foundation, Evolution, Mechanism, Ability, and Outlookcited by 43
  9. 2025VisMem: Latent Vision Memory Unlocks Potential of Vision-Language Modelscited by 37
  10. 2025Open Vision Reasoner: Transferring Linguistic Cognitive Behavior for Visual Reasoningcited by 34
  11. 2025RealUnify: Do Unified Models Truly Benefit from Unification? A Comprehensive Benchmarkcited by 33
  12. 2025Multimodal Chain of Continuous Thought for Latent-Space Reasoning in Vision-Language Modelscited by 32
38 more
  1. 2026LaST0: Latent Spatio-Temporal Chain-of-Thought for Robotic Vision-Language-Action Modelcited by 31
  2. 2025Explain Before You Answer: A Survey on Compositional Visual Reasoningcited by 28
  3. 2025Reasoning Within the Mind: Dynamic Multimodal Interleaving in Latent Spacecited by 28
  4. 2025Sketch-in-Latents: Eliciting Unified Reasoning in MLLMscited by 25
  5. 2025Reasoning in the Dark: Interleaved Vision-Text Reasoning in Latent Spacecited by 23
  6. 2026Thinking with Blueprints: Assisting Vision-Language Models in Spatial Reasoning via Structured Object Representationcited by 21
  7. 2025Mull-Tokens: Modality-Agnostic Latent Thinkingcited by 19
  8. 2025DiffThinker: Towards Generative Multimodal Reasoning with Diffusion Modelscited by 19
  9. 2025Latent Implicit Visual Reasoningcited by 18
  10. 2026Forest Before Trees: Latent Superposition for Efficient Visual Reasoningcited by 18
  11. 2026Vision-aligned Latent Reasoning for Multi-modal Large Language Modelcited by 18
  12. 2026LaViT: Aligning Latent Visual Thoughts for Multi-modal Reasoningcited by 15
  13. 2025SpinBench: Perspective and Rotation as a Lens on Spatial Reasoning in VLMscited by 14
  14. 2025DeepSketcher: Internalizing Visual Manipulation for Multimodal Reasoningcited by 14
  15. 2026Imagination Helps Visual Reasoning, But Not Yet in Latent Spacecited by 12
  16. 2025SpatialDreamer: Incentivizing Spatial Reasoning via Active Mental Imagerycited by 11
  17. 2026Visually-Guided Policy Optimization for Multimodal Reasoningcited by 11
  18. 2026Residual Context Diffusion Language Modelscited by 10
  19. 2026Dual Latent Memory for Visual Multi-agent Systemcited by 10
  20. 2025Octopus: Agentic Multimodal Reasoning with Six-Capability Orchestrationcited by 9
  21. 2025COOPER: A Unified Model for Cooperative Perception and Reasoning in Spatial Intelligencecited by 9
  22. 2026Thinking Without Images: Internalizing Visual Manipulation with On-Policy Self-Distillationcited by 9
  23. 2025Small Drafts, Big Verdict: Information-Intensive Visual Reasoning via Speculationcited by 8
  24. 2025LaRe: Latent Refocusing for Multimodal Reasoningcited by 8
  25. 2026Show, Don't Tell: Morphing Latent Reasoning into Image Generationcited by 8
  26. 2026LatentUM: Unleashing the Potential of Interleaved Cross-Modal Reasoning via a Latent-Space Unified Modelcited by 8
  27. 2026LaST-R1: Reinforcing Robotic Manipulation via Adaptive Physical Latent Reasoningcited by 8
  28. 2026UniVLR: Unifying Text and Vision in Visual Latent Reasoning for Multimodal LLMscited by 8
  29. 2026Toward Cognitive Supersensing in Multimodal Large Language Modelcited by 7
  30. 2026MentisOculi: Revealing the Limits of Reasoning with Mental Imagerycited by 7
  31. 2026Learning Modal-Mixed Chain-of-Thought Reasoning with Latent Embeddingscited by 6
  32. 2026Towards Explainable Industrial Anomaly Detection via Knowledge-Guided Latent Reasoningcited by 6
  33. 2026CrystaL: Spontaneous Emergence of Visual Latents in MLLMscited by 6
  34. 2026LanteRn: Latent Visual Structured Reasoningcited by 6
  35. 2026HyLaR: Hybrid Latent Reasoning with Decoupled Policy Optimizationcited by 6
  36. 2026Visual Latents Know More Than They Say: Unsilencing Latent Reasoning in MLLMscited by 6
  37. 2026Retrieve, Integrate, and Synthesize: Spatial-Semantic Grounded Latent Visual Reasoningcited by 6
  38. 2026VBVR-Pro: A Scalable and Verifiable Suite for Native Visual Reasoningcited by 6

50 most cited of 124 All 124 (JSON)

Similar, not cited

Close in wording, no citation either way

  1. 2025See, Think, Learn: A Self-Taught Multimodal Reasoner
  2. 2026See It, Say It, Sorted: An Iterative Training-Free Framework for Visually-Grounded Multimodal Reasoning in LVLMs
  3. 2025From Perception to Cognition: A Survey of Vision-Language Interactive Reasoning in Multimodal Large Language Models
  4. 2025From Illusion to Intention: Visual Rationale Learning for Vision-Language Reasoning
  5. 2025Thinking with Images for Multimodal Reasoning: Foundations, Methods, and Future Frontiers
  6. 2026Retrieved Images as Visual Thought: Training-Free Multimodal In-Context Learning for the Open-vs-Closed Gap
  7. 2025Beyond Seeing: Evaluating Multimodal LLMs on Tool-Enabled Image Perception, Transformation, and Reasoning
  8. 2025VERIFY: A Benchmark of Visual Explanation and Reasoning for Investigating Multimodal Reasoning Fidelity