Circular Reasoning: Understanding Self-Reinforcing Loops in Large Reasoning Models

TL;DR

This paper investigates the phenomenon of circular reasoning in large reasoning models, where generated content self-reinforces, causing loops that hinder inference, and introduces methods to analyze and predict these loops.

Contribution

The paper identifies circular reasoning as a distinct failure mode in LRMs, introduces LoopBench dataset, and develops CUSUM-based early prediction for these self-reinforcing loops.

Findings

Circular reasoning manifests as semantic and textual repetition in LRMs.

Reasoning impasses trigger and sustain self-reinforcing loops.

CUSUM algorithm effectively predicts loop onset in diverse models.

Abstract

Despite the success of test-time scaling, Large Reasoning Models (LRMs) frequently encounter repetitive loops that lead to computational waste and inference failure. In this paper, we identify a distinct failure mode termed Circular Reasoning. Unlike traditional model degeneration, this phenomenon manifests as a self-reinforcing trap where generated content acts as a logical premise for its own recurrence, compelling the reiteration of preceding text. To systematically analyze this phenomenon, we introduce LoopBench, a dataset designed to capture two distinct loop typologies: numerical loops and statement loops. Mechanistically, we characterize circular reasoning as a state collapse exhibiting distinct boundaries, where semantic repetition precedes textual repetition. We reveal that reasoning impasses trigger the loop onset, which subsequently persists as an inescapable cycle driven by a self-reinforcing V-shaped attention mechanism. Guided by these findings, we employ the Cumulative Sum (CUSUM) algorithm to capture these precursors for early loop prediction. Experiments across diverse LRMs validate its accuracy and elucidate the stability of long-chain reasoning.