Stop When Reasoning Converges: Semantic-Preserving Early Exit for Reasoning Models

TL;DR

This paper introduces PUMA, a framework that uses reasoning-level semantic redundancy to enable early exits in reasoning models, reducing tokens and latency without sacrificing accuracy.

Contribution

It proposes a novel semantic redundancy signal for early stopping in reasoning models and develops PUMA, a plug-and-play framework combining redundancy detection and verification.

Findings

PUMA reduces 26.2% of tokens on average across benchmarks.

Semantic redundancy effectively indicates reasoning convergence.

PUMA maintains accuracy while enabling early exits in various tasks.

Abstract

Large Reasoning Models (LRMs) achieve strong performance by generating long chains of thought (CoT), but often overthink, continuing to reason after a solution has already stabilized and thereby wasting tokens and increasing latency. Existing inference-time early-exit methods rely primarily on answer-level signals, such as confidence or trial-answer consistency, to decide when to stop. However, these signals mainly reflect answer readiness rather than reasoning convergence: they may trigger before the model has finished exploring or self-correcting, causing premature exits that can degrade final-answer accuracy and leave the retained reasoning chain semantically incomplete. We identify reasoning-level semantic redundancy as a complementary signal for semantic-preserving early exit: when successive steps no longer add novel progress and instead revisit established conclusions, the reasoning trajectory has likely converged. Building on this insight, we propose PUMA, a plug-and-play framework that combines a lightweight Redundancy Detector with answer-level verification. The detector flags semantically redundant candidate exits, while verification confirms whether stopping is safe, allowing PUMA to remove redundant continuation while preserving both answer accuracy and a coherent reasoning prefix. Across five LRMs and five challenging reasoning benchmarks, PUMA achieves 26.2% average token reduction while preserving accuracy and retained CoT quality. Additional experiments on code generation, zero-shot vision-language reasoning, and learned stopping-policy internalization further demonstrate that reasoning-level redundancy is a robust, transferable, and learnable signal for efficient reasoning. Our code is available at \url{https://github.com/giovanni-vaccarino/PUMA}.