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

TL;DR

This paper identifies hallucinations in multimodal reasoning models linked to high-uncertainty cognitive bifurcation points and proposes V-STAR, a training paradigm with attention reinforcement and reflection mechanisms to improve visual grounding.

Contribution

It introduces V-STAR, a novel training framework with hierarchical attention rewards and reflection strategies to mitigate hallucinations in multimodal reasoning models.

Findings

V-STAR improves visual grounding during high-uncertainty reasoning.

The hierarchical attention reward enhances visual evidence querying.

Reflection mechanisms reduce hallucination frequency.

Abstract

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.