The Cost of Reasoning: Chain-of-Thought Induces Overconfidence in Vision-Language Models

TL;DR

This paper investigates how chain-of-thought reasoning in vision-language models affects uncertainty quantification, revealing that reasoning often leads to overconfidence due to implicit answer conditioning, despite improving accuracy.

Contribution

It uncovers the impact of reasoning on uncertainty estimates in VLMs and identifies implicit answer conditioning as a key factor causing overconfidence.

Findings

Reasoning degrades uncertainty estimate quality.

Implicit answer conditioning causes overconfidence.

Agreement-based consistency remains reliable for UQ.

Abstract

Vision-language models (VLMs) are increasingly deployed in high-stakes settings where reliable uncertainty quantification (UQ) is as important as predictive accuracy. Extended reasoning via chain-of-thought (CoT) prompting or reasoning-trained models has become ubiquitous in modern VLM pipelines, yet its effect on UQ reliability remains poorly understood. We show that reasoning consistently degrades the quality of most uncertainty estimates, even when it improves task accuracy. We identify implicit answer conditioning as the primary mechanism: as reasoning traces converge on a conclusion before the final answer is generated, token probabilities increasingly reflect consistency with the model's own reasoning trace rather than uncertainty about correctness. In effect, the model becomes overconfident in its answer. In contrast, agreement-based consistency remains robust and often improves under reasoning, making it a practical choice for uncertainty estimation in reasoning-enabled VLMs.