mlsanityData to 7 Oct 2026

Cognitive Pivot Points and Visual Anchoring: Unveiling and Rectifying Hallucinations in Multimodal Reasoning Models

arXiv 2604.10219 ↗First posted 11 Apr 2026cs.AIcited by 1

On the map: LLM reasoning → Multimodal reasoning

Abstract (arXiv)

Multimodal Large Reasoning Models (MLRMs) have achieved remarkable strides in visual reasoning through test time compute scaling, yet long chain reasoning remains prone to hallucinations. We identify a concerning phenomenon termed the Reasoning Vision Truth Disconnect (RVTD): hallucinations are strongly correlated with cognitive bifurcation points that often exhibit high entropy states. We attribute this vulnerability to a breakdown in visual semantic anchoring, localized within the network's intermediate layers; specifically, during these high uncertainty transitions, the model fails to query visual evidence, reverting instead to language priors. Consequently, we advocate a shift from solely outcome level supervision to augmenting it with fine grained internal attention guidance. To this end, we propose V-STAR (Visual Structural Training with Attention Reinforcement), a lightweight, holistic training paradigm designed to internalize visually aware reasoning capabilities. Central to our approach is the Hierarchical Visual Attention Reward (HVAR), integrated within the GRPO framework. Upon detecting high entropy states, this mechanism dynamically incentivizes visual attention across critical intermediate layers, thereby anchoring the reasoning process back to the visual input. Furthermore, we introduce the Forced Reflection Mechanism (FRM), a trajectory editing strategy that disrupts cognitive inertia by triggering reflection around high entropy cognitive bifurcation points and encouraging verification of subsequent steps against the visual input, thereby translating external debiasing interventions into an intrinsic capability for hallucination mitigation.

Key facts. Cited by 1 paper in the map (data through 7 Oct 2026). Placed in Multimodal reasoning, one of 17 directions in LLM reasoning. Builds on 57 of its 106 references that are in the map. First posted on arXiv 11 Apr 2026.

Builds on

57 of 106 references.

Closest ancestors · 14

Same direction, newest first

  1. 2025From Perception to Cognition: A Survey of Vision-Language Interactive Reasoning in Multimodal Large Language Models29 Sep
  2. 2025VL-Cogito: Progressive Curriculum Reinforcement Learning for Advanced Multimodal Reasoning30 Jul
  3. 2025VisuLogic: A Benchmark for Evaluating Visual Reasoning in Multi-modal Large Language Models21 Apr
  4. 2025VL-Rethinker: Incentivizing Self-Reflection of Vision-Language Models with Reinforcement Learning10 Apr
  5. 2025SoTA with Less: MCTS-Guided Sample Selection for Data-Efficient Visual Reasoning Self-Improvement10 Apr
  6. 2025OpenVLThinker: Complex Vision-Language Reasoning via Iterative SFT-RL Cycles21 Mar
  7. 2025R1-Onevision: Advancing Generalized Multimodal Reasoning through Cross-Modal Formalization13 Mar
  8. 2025MM-Eureka: Exploring the Frontiers of Multimodal Reasoning with Rule-based Reinforcement Learning10 Mar
  9. 2025Vision-R1: Incentivizing Reasoning Capability in Multimodal Large Language Models9 Mar
  10. 2024MMEvalPro: Calibrating Multimodal Benchmarks Towards Trustworthy and Efficient Evaluation29 Jun
  11. 2024MM-MATH: Advancing Multimodal Math Evaluation with Process Evaluation and Fine-grained Classification7 Apr
  12. 2024Measuring Multimodal Mathematical Reasoning with MATH-Vision Dataset22 Feb
2 more
  1. 2023MathVista: Evaluating Mathematical Reasoning of Foundation Models in Visual Contexts3 Oct
  2. 2023Multimodal Chain-of-Thought Reasoning in Language Models2 Feb

Other references · 43

  1. 2017Proximal Policy Optimization Algorithmscited by 14,441
  2. 2024The Llama 3 Herd of Modelscited by 14,192
  3. 2024DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Modelscited by 8,851
  4. 2025DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learningcited by 6,146
  5. 2022Self-Consistency Improves Chain of Thought Reasoning in Language Modelscited by 5,601
  6. 2025Qwen2.5-VL Technical Reportcited by 5,418
  7. 2024GPT-4o System Cardcited by 4,467
  8. 2024Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolutioncited by 4,151
  9. 2023MiniGPT-4: Enhancing Vision-Language Understanding with Advanced Large Language Modelscited by 2,689
  10. 2025DAPO: An Open-Source LLM Reinforcement Learning System at Scalecited by 2,678
  11. 2023Sparks of Artificial General Intelligence: Early experiments with GPT-4cited by 2,579
  12. 2015High-Dimensional Continuous Control Using Generalized Advantage Estimationcited by 1,916
31 more
  1. 2023Evaluating Object Hallucination in Large Vision-Language Modelscited by 1,696
  2. 2016Concrete Problems in AI Safetycited by 1,582
  3. 2025Understanding R1-Zero-Like Training: A Critical Perspectivecited by 1,276
  4. 2023MM-Vet: Evaluating Large Multimodal Models for Integrated Capabilitiescited by 1,221
  5. 2024OpenAI o1 System Cardcited by 1,220
  6. 2024Qwen2.5-Math Technical Report: Toward Mathematical Expert Model via Self-Improvementcited by 1,057
  7. 2022In-context Learning and Induction Headscited by 930
  8. 2022Toy Models of Superpositioncited by 928
  9. 2024Are We on the Right Way for Evaluating Large Vision-Language Models?cited by 893
  10. 2023Shikra: Unleashing Multimodal LLM's Referential Dialogue Magiccited by 891
  11. 2024Eyes Wide Shut? Exploring the Visual Shortcomings of Multimodal LLMscited by 719
  12. 2023Aligning Large Multimodal Models with Factually Augmented RLHFcited by 703
  13. 2023The Geometry of Truth: Emergent Linear Structure in Large Language Model Representations of True/False Datasetscited by 673
  14. 2024MathVerse: Does Your Multi-modal LLM Truly See the Diagrams in Visual Math Problems?cited by 573
  15. 2023HallusionBench: An Advanced Diagnostic Suite for Entangled Language Hallucination and Visual Illusion in Large Vision-Language Modelscited by 572
  16. 2025SimpleRL-Zoo: Investigating and Taming Zero Reinforcement Learning for Open Base Models in the Wildcited by 505
  17. 2023V*: Guided Visual Search as a Core Mechanism in Multimodal LLMscited by 459
  18. 2023Mitigating Hallucination in Large Multi-Modal Models via Robust Instruction Tuningcited by 439
  19. 2024Hallucination of Multimodal Large Language Models: A Surveycited by 380
  20. 2021Inter-GPS: Interpretable Geometry Problem Solving with Formal Language and Symbolic Reasoningcited by 376
  21. 2023DoLa: Decoding by Contrasting Layers Improves Factuality in Large Language Modelscited by 367
  22. 2023SPHINX: The Joint Mixing of Weights, Tasks, and Visual Embeddings for Multi-modal Large Language Modelscited by 283
  23. 2023Holistic Analysis of Hallucination in GPT-4V(ision): Bias and Interference Challengescited by 131
  24. 2025ZeroBench: An Impossible Visual Benchmark for Contemporary Large Multimodal Modelscited by 56
  25. 2025PhyX: Does Your Model Have the "Wits" for Physical Reasoning?cited by 37
  26. 2025Automated Generation of Challenging Multiple-Choice Questions for Vision Language Model Evaluationcited by 28
  27. 2025PhysUniBench: A Multi-Modal Physics Reasoning Benchmark at Undergraduate Levelcited by 14
  28. 2026Thinking in Uncertainty: Mitigating Hallucinations in MLRMs with Latent Entropy-Aware Decodingcited by 11
  29. 2025Scaling Up RL: Unlocking Diverse Reasoning in LLMs via Prolonged Trainingcited by 5
  30. 2025Geometric Prior-Guided Federated Prompt Calibrationcited by 2
  31. 2026Region-Constrained Group Relative Policy Optimization for Flow-Based Image Editingcited by 2

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 1.

1 paper.

  1. 2026LatentOmni: Rethinking Omni-Modal Understanding via Unified Audio-Visual Latent Reasoningcited by 5

Similar, not cited

Close in wording, no citation either way

  1. 2025Reallocating Attention Across Layers to Reduce Multimodal Hallucination
  2. 2026Seeing Through the Chain: Mitigate Hallucination in Multimodal Reasoning Models via CoT Compression and Contrastive Preference Optimization
  3. 2024Thinking Before Looking: Improving Multimodal LLM Reasoning via Mitigating Visual Hallucination
  4. 2025More Thinking, Less Seeing? Assessing Amplified Hallucination in Multimodal Reasoning Models
  5. 2026Understanding the Role of Hallucination in Reinforcement Post-Training of Multimodal Reasoning Models
  6. 2026ClueTracer: Question-to-Vision Clue Tracing for Training-Free Hallucination Suppression in Multimodal Reasoning
  7. 2025Visual Attention Reasoning via Hierarchical Search and Self-Verification
  8. 2026See It, Say It, Sorted: An Iterative Training-Free Framework for Visually-Grounded Multimodal Reasoning in LVLMs