mlsanityData to 14 Oct 2026

Linking Perception, Confidence and Accuracy in MLLMs

arXiv 2603.12149 ↗First posted 12 Mar 2026cs.CV cs.CLcited by 2

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

Abstract (arXiv)

Recent advances in Multi-modal Large Language Models (MLLMs) have predominantly focused on enhancing visual perception to improve accuracy. However, a critical question remains unexplored: Do models know when they do not know? Through a probing experiment, we reveal a severe confidence miscalibration problem in MLLMs. To address this, we propose Confidence-Driven Reinforcement Learning (CDRL), which uses original-noise image pairs and a novel confidence-based reward to enhance perceptual sensitivity and robustly calibrate the model's confidence. Beyond training benefits, calibrated confidence enables more effective test-time scaling as a free lunch. We further propose Confidence-Aware Test-Time Scaling (CA-TTS), which dynamically coordinates Self-Consistency, Self-Reflection, and Visual Self-Check modules guided by confidence signals. An Expert Model acts in multiple roles (e.g., Planner, Critic, Voter) to schedule these modules and provide external verification. Our integrated framework establishes new state-of-the-art results with consistent 8.8% gains across four benchmarks. More ablation studies demonstrate the effectiveness of each module and scaling superiority.

Key facts. Cited by 2 papers in the map (data through 14 Oct 2026). Placed in Hallucination mitigation in LVLMs, one of 16 directions in Multimodal LLMs / vision-language models. Builds on 49 of its 57 references that are in the map. First posted on arXiv 12 Mar 2026.

Builds on

49 of 57 references.

Closest ancestors · 7

Same direction, newest first

  1. 2025Mitigating Hallucinations in Multimodal LLMs via Object-aware Preference Optimization27 Aug
  2. 2025Mitigating Hallucinations in Large Vision-Language Models via DPO: On-Policy Data Hold the Key16 Jan
  3. 2024Mitigating Hallucination in Multimodal Large Language Model via Hallucination-targeted Direct Preference Optimization15 Nov
  4. 2024V-DPO: Mitigating Hallucination in Large Vision Language Models via Vision-Guided Direct Preference Optimization5 Nov
  5. 2024ConVis: Contrastive Decoding with Hallucination Visualization for Mitigating Hallucinations in Multimodal Large Language Models25 Aug
  6. 2023RLHF-V: Towards Trustworthy MLLMs via Behavior Alignment from Fine-grained Correctional Human Feedback1 Dec
  7. 2023Mitigating Object Hallucinations in Large Vision-Language Models through Visual Contrastive Decoding28 Nov

Other references · 42

  1. 2022Chain-of-Thought Prompting Elicits Reasoning in Large Language Modelscited by 12,459
  2. 2023Visual Instruction Tuningcited by 7,869
  3. 2025Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilitiescited by 3,796
  4. 2023Tree of Thoughts: Deliberate Problem Solving with Large Language Modelscited by 3,211
  5. 2023MMBench: Is Your Multi-modal Model an All-around Player?cited by 2,141
  6. 2023MMMU: A Massive Multi-discipline Multimodal Understanding and Reasoning Benchmark for Expert AGIcited by 2,080
  7. 2023MathVista: Evaluating Mathematical Reasoning of Foundation Models in Visual Contextscited by 1,614
  8. 2023Can LLMs Express Their Uncertainty? An Empirical Evaluation of Confidence Elicitation in LLMscited by 974
  9. 2024Are We on the Right Way for Evaluating Large Vision-Language Models?cited by 914
  10. 2023Just Ask for Calibration: Strategies for Eliciting Calibrated Confidence Scores from Language Models Fine-Tuned with Human Feedbackcited by 812
  11. 2024MathVerse: Does Your Multi-modal LLM Truly See the Diagrams in Visual Math Problems?cited by 586
  12. 2024Measuring Multimodal Mathematical Reasoning with MATH-Vision Datasetcited by 526
30 more
  1. 2024MMMU-Pro: A More Robust Multi-discipline Multimodal Understanding Benchmarkcited by 512
  2. 2023Trustworthy LLMs: a Survey and Guideline for Evaluating Large Language Models' Alignmentcited by 401
  3. 2025R1-Onevision: Advancing Generalized Multimodal Reasoning through Cross-Modal Formalizationcited by 370
  4. 2025VL-Rethinker: Incentivizing Self-Reflection of Vision-Language Models with Reinforcement Learningcited by 323
  5. 2023A Survey of Confidence Estimation and Calibration in Large Language Modelscited by 256
  6. 2024LogicVista: Multimodal LLM Logical Reasoning Benchmark in Visual Contextscited by 245
  7. 2024Enhancing the Reasoning Ability of Multimodal Large Language Models via Mixed Preference Optimizationcited by 239
  8. 2025A Survey on Test-Time Scaling in Large Language Models: What, How, Where, and How Well?cited by 223
  9. 2025Deep Think with Confidencecited by 205
  10. 2025Scalable Best-of-N Selection for Large Language Models via Self-Certaintycited by 187
  11. 2024DynaMath: A Dynamic Visual Benchmark for Evaluating Mathematical Reasoning Robustness of Vision Language Modelscited by 146
  12. 2025S*: Test Time Scaling for Code Generationcited by 98
  13. 2024When Do LLMs Need Retrieval Augmentation? Mitigating LLMs' Overconfidence Helps Retrieval Augmentationcited by 59
  14. 2025Calibrating Verbal Uncertainty as a Linear Feature to Reduce Hallucinationscited by 48
  15. 2024When to Trust LLMs: Aligning Confidence with Response Qualitycited by 43
  16. 2025A Survey of Test-Time Compute: From Intuitive Inference to Deliberate Reasoningcited by 40
  17. 2025Reasoning on a Budget: A Survey of Adaptive and Controllable Test-Time Compute in LLMscited by 31
  18. 2025DepthLM: Metric Depth From Vision Language Modelscited by 30
  19. 2025Learning to See Before Seeing: Demystifying LLM Visual Priors from Language Pre-trainingcited by 23
  20. 2023Eliciting In-Context Learning in Vision-Language Models for Videos Through Curated Data Distributional Propertiescited by 21
  21. 2025To Sink or Not to Sink: Visual Information Pathways in Large Vision-Language Modelscited by 20
  22. 2025Revisiting Multimodal Positional Encoding in Vision-Language Modelscited by 16
  23. 2025DreamPRM: Domain-Reweighted Process Reward Model for Multimodal Reasoningcited by 15
  24. 2025Revisiting Self-Consistency from Dynamic Distributional Alignment Perspective on Answer Aggregationcited by 11
  25. 2025CMCTS: A Constrained Monte Carlo Tree Search Framework for Mathematical Reasoning in Large Language Modelcited by 9
  26. 2025Prompt4Trust: A Reinforcement Learning Prompt Augmentation Framework for Clinically-Aligned Confidence Calibration in Multimodal Large Language Modelscited by 6
  27. 2025Exposing and Mitigating Calibration Biases and Demographic Unfairness in MLLM Few-Shot In-Context Learning for Medical Image Classificationcited by 5
  28. 2025What do vision-language models see in the context? Investigating multimodal in-context learningcited by 4
  29. 2025RL makes MLLMs see better than SFTcited by 3
  30. 2025Test-Time Scaling Strategies for Generative Retrieval in Multimodal Conversational Recommendationscited by 1

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 2.

2 papers.

  1. 2026Evidence-RL: Towards Evidence-intensive Visual Reasoningcited by 3
  2. 2026The Mirage of Calibrated Confidence: Trajectory-Independence of Verbalized Confidence in Vision-Language Modelscited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026VL-Calibration: Decoupled Confidence Calibration for Large Vision-Language Models Reasoning
  2. 2026Scaling with Confidence: Calibrating Confidence of LLMs for Adaptive Test Time Scaling
  3. 2026Instinct vs. Reflection: Unifying Token and Verbalized Confidence in Multimodal Large Models
  4. 2025Seeing is Believing, but How Much? A Comprehensive Analysis of Verbalized Calibration in Vision-Language Models
  5. 2025MMBoundary: Advancing MLLM Knowledge Boundary Awareness through Reasoning Step Confidence Calibration
  6. 2026Ranking-Aware Calibration for Reliable Multimodal Reinforcement Learning
  7. 2026Shared Experience, Separate Learning: Companion Confidence Calibration for LLMs
  8. 2025Rewarding Doubt: A Reinforcement Learning Approach to Calibrated Confidence Expression of Large Language Models