Analyzing Reasoning Consistency in Large Multimodal Models under Cross-Modal Conflicts

TL;DR

This paper investigates the reasoning robustness of large multimodal models, revealing their tendency to propagate textual hallucinations and proposing a visual grounding method to improve their accuracy and reliability.

Contribution

It introduces the LogicGraph Perturbation Protocol for evaluating reasoning robustness and proposes Active Visual-Context Refinement to mitigate hallucinations without retraining.

Findings

Models succeed in self-correction in less than 10% of cases

Textual hallucinations tend to propagate blindly in reasoning chains

Proposed method significantly reduces hallucination propagation and improves robustness

Abstract

Large Multimodal Models (LMMs) have demonstrated impressive capabilities in video reasoning via Chain-of-Thought (CoT). However, the robustness of their reasoning chains remains questionable. In this paper, we identify a critical failure mode termed textual inertia, where once a textual hallucination occurs in the thinking process, models tend to blindly adhere to the erroneous text while neglecting conflicting visual evidence. To systematically investigate this, we propose the LogicGraph Perturbation Protocol that structurally injects perturbations into the reasoning chains of diverse LMMs spanning both native reasoning architectures and prompt-driven paradigms to evaluate their self-reflection capabilities. The results reveal that models successfully self-correct in less than 10% of cases and predominantly succumb to blind textual error propagation. To mitigate this, we introduce Active Visual-Context Refinement, a training-free inference paradigm which orchestrates an active visual re-grounding mechanism to enforce fine-grained verification coupled with an adaptive context refinement strategy to summarize and denoise the reasoning history. Experiments demonstrate that our approach significantly stifles hallucination propagation and enhances reasoning robustness.