mlsanityData to 7 Oct 2026

Grounding the Score: Explicit Visual Premise Verification for Reliable Vision-Language Process Reward Models

arXiv 2603.16253 ↗First posted 17 Mar 2026cs.CV cs.AIcited by 4

On the map: LLM reasoning → Multimodal reasoning

also in 1 more
Abstract (arXiv)

Vision-language process reward models (VL-PRMs) are increasingly used to score intermediate reasoning steps and rerank candidates under test-time scaling. However, they often function as black-box judges: a low step score may reflect a genuine reasoning mistake or simply the verifier's misperception of the image. This entanglement between perception and reasoning leads to systematic false positives (rewarding hallucinated visual premises) and false negatives (penalizing correct grounded statements), undermining both reranking and error localization. We introduce Explicit Visual Premise Verification (EVPV), a lightweight verification interface that conditions step scoring on the reliability of the visual premises a step depends on. The policy is prompted to produce a step-wise visual checklist that makes required visual facts explicit, while a constraint extractor independently derives structured visual constraints from the input image. EVPV matches checklist claims against these constraints to compute a scalar visual reliability signal, and calibrates PRM step rewards via reliability gating: rewards for visually dependent steps are attenuated when reliability is low and preserved when reliability is high. This decouples perceptual uncertainty from logical evaluation without per-step tool calls. Experiments on VisualProcessBench and six multimodal reasoning benchmarks show that EVPV improves step-level verification and consistently boosts Best-of-N reranking accuracy over strong baselines. Furthermore, injecting controlled corruption into the extracted constraints produces monotonic performance degradation, providing causal evidence that the gains arise from constraint fidelity and explicit premise verification rather than incidental prompt effects. Code is available at: https://github.com/Qwen-Applications/EVPV-PRM

Key facts. Cited by 4 papers in the map (data through 7 Oct 2026). Placed in Multimodal reasoning, one of 17 directions in LLM reasoning. Builds on 50 of its 53 references that are in the map. First posted on arXiv 17 Mar 2026.

Builds on

50 of 53 references.

Closest ancestors · 12

Same direction, newest first

  1. 2025MM-CoT:A Benchmark for Probing Visual Chain-of-Thought Reasoning in Multimodal Models9 Dec
  2. 2025TIM-PRM: Verifying multimodal reasoning with Tool-Integrated PRM28 Nov
  3. 2025Training Vision-Language Process Reward Models for Test-Time Scaling in Multimodal Reasoning: Key Insights and Lessons Learned27 Sep
  4. 2025GM-PRM: A Generative Multimodal Process Reward Model for Multimodal Mathematical Reasoning6 Aug
  5. 2025DreamPRM: Domain-Reweighted Process Reward Model for Multimodal Reasoning26 May
  6. 2025VisualPRM: An Effective Process Reward Model for Multimodal Reasoning13 Mar
  7. 2025VLRMBench: A Comprehensive and Challenging Benchmark for Vision-Language Reward Models10 Mar
  8. 2025Unlocking Multimodal Mathematical Reasoning via Process Reward Model8 Jan
  9. 2024LogicVista: Multimodal LLM Logical Reasoning Benchmark in Visual Contexts6 Jul
  10. 2024We-Math: Does Your Large Multimodal Model Achieve Human-like Mathematical Reasoning?1 Jul
  11. 2024Measuring Multimodal Mathematical Reasoning with MATH-Vision Dataset22 Feb
  12. 2023MathVista: Evaluating Mathematical Reasoning of Foundation Models in Visual Contexts3 Oct

Other references · 38

  1. 2023MMMU: A Massive Multi-discipline Multimodal Understanding and Reasoning Benchmark for Expert AGIcited by 2,044
  2. 2025InternVL3: Exploring Advanced Training and Test-Time Recipes for Open-Source Multimodal Modelscited by 1,769
  3. 2024MathVerse: Does Your Multi-modal LLM Truly See the Diagrams in Visual Math Problems?cited by 573
  4. 2023Language Is Not All You Need: Aligning Perception with Language Modelscited by 543
  5. 2024BLINK: Multimodal Large Language Models Can See but Not Perceivecited by 529
  6. 2025The Lessons of Developing Process Reward Models in Mathematical Reasoningcited by 402
  7. 2024ReST-MCTS*: LLM Self-Training via Process Reward Guided Tree Searchcited by 338
  8. 2024VisionLLM v2: An End-to-End Generalist Multimodal Large Language Model for Hundreds of Vision-Language Taskscited by 136
  9. 2025Process Reward Models That Thinkcited by 118
  10. 2025PRMBench: A Fine-grained and Challenging Benchmark for Process-Level Reward Modelscited by 98
  11. 2023Let's reward step by step: Step-Level reward model as the Navigators for Reasoningcited by 96
  12. 2025GenPRM: Scaling Test-Time Compute of Process Reward Models via Generative Reasoningcited by 94
26 more
  1. 2024Process Reward Model with Q-Value Rankingscited by 78
  2. 2025AgentPRM: Process Reward Models for LLM Agents via Step-Wise Promise and Progresscited by 69
  3. 2025Stop Summation: Min-Form Credit Assignment Is All Process Reward Model Needs for Reasoningcited by 49
  4. 2025Reward Reasoning Modelcited by 47
  5. 2025Writing-Zero: Bridge the Gap Between Non-verifiable Tasks and Verifiable Rewardscited by 47
  6. 2023VCoder: Versatile Vision Encoders for Multimodal Large Language Modelscited by 42
  7. 2025VersaPRM: Multi-Domain Process Reward Model via Synthetic Reasoning Datacited by 39
  8. 2023Visual cognition in multimodal large language modelscited by 38
  9. 2024Entropy-Regularized Process Reward Modelcited by 38
  10. 2025R-PRM: Reasoning-Driven Process Reward Modelingcited by 38
  11. 2025A Survey of Process Reward Models: From Outcome Signals to Process Supervisions for Large Language Modelscited by 37
  12. 2025Generate, but Verify: Reducing Hallucination in Vision-Language Models with Retrospective Resamplingcited by 32
  13. 2024TextHawk: Exploring Efficient Fine-Grained Perception of Multimodal Large Language Modelscited by 27
  14. 2025Introducing Visual Perception Token into Multimodal Large Language Modelcited by 26
  15. 2026Open Rubric System: Scaling Reinforcement Learning with Pairwise Adaptive Rubriccited by 26
  16. 2023Chain of Visual Perception: Harnessing Multimodal Large Language Models for Zero-shot Camouflaged Object Detectioncited by 22
  17. 2025AURORA:Automated Training Framework of Universal Process Reward Models via Ensemble Prompting and Reverse Verificationcited by 22
  18. 2025Uncertainty-Aware Step-wise Verification with Generative Reward Modelscited by 21
  19. 2025Know What You Don't Know: Uncertainty Calibration of Process Reward Modelscited by 19
  20. 2025Athena: Enhancing Multimodal Reasoning with Data-efficient Process Reward Modelscited by 16
  21. 2025Dynamic and Generalizable Process Reward Modelingcited by 16
  22. 2025Efficient Process Reward Model Training via Active Learningcited by 14
  23. 2025VRPRM: Process Reward Modeling via Visual Reasoningcited by 11
  24. 2025Self-Guided Process Reward Optimization with Redefined Step-wise Advantage for Process Reinforcement Learningcited by 10
  25. 2025Error Typing for Smarter Rewards: Improving Process Reward Models with Error-Aware Hierarchical Supervisioncited by 8
  26. 2025The Bidirectional Process Reward Modelcited by 4

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 4.

4 papers.

  1. 2026Verification Mirage: Mapping the Reliability Boundary of Self-Verification in Medical VQAcited by 1
  2. 2026Best-of-Evidence: Best-of-N Selection under Partial Verificationcited by 0
  3. 2026Separating perception from reasoning in vision-language models: a model-free render ceiling for crystal structurescited by 0
  4. 2026A Low Grounding Score Is Not an Ungrounded Judge: Identifying the Perceptibility Confound in Multimodal Oversightcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026ReGround: Restoring Visual Grounding in Multi-Step Reasoning through Self-Diagnosis and Visual Re-Examination
  2. 2026Step-Level Visual Grounding Faithfulness Predicts Out-of-Distribution Generalization in Long-Horizon Vision-Language Models
  3. 2026What, Whether and How? Unveiling Process Reward Models for Thinking with Images Reasoning
  4. 2025VERIFY: A Benchmark of Visual Explanation and Reasoning for Investigating Multimodal Reasoning Fidelity
  5. 2025TRACE: A Framework for Analyzing and Enhancing Stepwise Reasoning in Vision-Language Models
  6. 2025BLINK-Twice: You see, but do you observe? A Reasoning Benchmark on Visual Perception
  7. 2025Journey Before Destination: On the importance of Visual Faithfulness in Slow Thinking
  8. 2026ComplexityWorld: Benchmarking Vision-Language Models on Verifiable Visual Decision Making