Thinkless: LLM Learns When to Think

TL;DR

Thinkless enables LLMs to adaptively choose between concise and detailed reasoning modes, significantly improving efficiency by reducing unnecessary complex reasoning without sacrificing accuracy.

Contribution

This paper introduces a novel reinforcement learning framework with DeGRPO algorithm for LLMs to learn when to think, balancing reasoning depth and computational efficiency.

Findings

Reduces long-chain reasoning by 50-90% on benchmarks.

Maintains high accuracy while improving efficiency.

Demonstrates effective control over reasoning mode selection.

Abstract

Reasoning Language Models, capable of extended chain-of-thought reasoning, have demonstrated remarkable performance on tasks requiring complex logical inference. However, applying elaborate reasoning for all queries often results in substantial computational inefficiencies, particularly when many problems admit straightforward solutions. This motivates an open question: Can LLMs learn when to think? To answer this, we propose Thinkless, a learnable framework that empowers an LLM to adaptively select between short-form and long-form reasoning, based on both task complexity and the model's ability. Thinkless is trained under a reinforcement learning paradigm and employs two control tokens, <short> for concise responses and <think> for detailed reasoning. At the core of our method is a Decoupled Group Relative Policy Optimization (DeGRPO) algorithm, which decomposes the learning objective of hybrid reasoning into two components: (1) a control token loss that governs the selection of the reasoning mode, and (2) a response loss that improves the accuracy of the generated answers. This decoupled formulation enables fine-grained control over the contributions of each objective, stabilizing training and effectively preventing collapse observed in vanilla GRPO. Empirically, on several benchmarks such as Minerva Algebra, MATH-500, and GSM8K, Thinkless is able to reduce the usage of long-chain thinking by 50% - 90%, significantly improving the efficiency of Reasoning Language Models. The code is available at https://github.com/VainF/Thinkless