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

2026-03-17

AI Summary

This work addresses the challenge that existing vision-language process reward models struggle to distinguish between reasoning errors and visual perception ambiguities, often misjudging hallucinated premises or penalizing valid image-dependent reasoning. To resolve this, the authors propose an Explicit Visual Premise Verification (EVPV) mechanism that generates a visual checklist and aligns it with structured constraints extracted from the image, thereby applying reliability-gated calibration to vision-dependent reasoning steps. EVPV is the first approach to explicitly integrate visual premise verification into process reward modeling, effectively disentangling perceptual uncertainty from logical evaluation without relying on external tools. Experiments demonstrate that EVPV significantly improves step-level verification accuracy and Best-of-N reranking performance on VisualProcessBench and six additional benchmarks, with controlled perturbation studies confirming that its gains stem from high-fidelity constraint matching.

Technology Category

• Computer Vision: Language and Vision
• Machine Learning: Calibration & Uncertainty Quantification
• Knowledge Representation and Reasoning: Reasoning with Beliefs

Application Category

• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work

Abstract

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

Problem

• vision-language process reward models
• visual premise verification
• false positives
• false negatives
• perceptual uncertainty

Innovation

• Explicit Visual Premise Verification
• Vision-Language Process Reward Models
• Reliability Gating
• Visual Constraint Extraction
• Step-wise Verification