From Efficiency to Adaptivity: A Deeper Look at Adaptive Reasoning in Large Language Models
Chao Wu, Baoheng Li, Mingchen Gao, Zhenyi Wang
TL;DR
This work reframes reasoning in large language models from a sole efficiency lens to adaptive reasoning, defined as input-dependent allocation of reasoning effort based on difficulty and uncertainty. It formalizes deductive, inductive, and abductive reasoning within LLMs and casts adaptive reasoning as a control-augmented policy optimization problem that balances task performance with compute cost. A comprehensive taxonomy separates training-based approaches (reinforcement learning, supervised fine-tuning, and learned controllers) from training-free strategies (prompt conditioning, inference-time halting, and modular composition), enabling systematic comparison across diverse mechanisms. The paper discusses open challenges in self-evaluation, meta-reasoning, and human-aligned control, and highlights practical implications for building more efficient, flexible, and context-sensitive reasoning systems.
Abstract
Recent advances in large language models (LLMs) have made reasoning a central benchmark for evaluating intelligence. While prior surveys focus on efficiency by examining how to shorten reasoning chains or reduce computation, this view overlooks a fundamental challenge: current LLMs apply uniform reasoning strategies regardless of task complexity, generating long traces for trivial problems while failing to extend reasoning for difficult tasks. This survey reframes reasoning through the lens of {adaptivity}: the capability to allocate reasoning effort based on input characteristics such as difficulty and uncertainty. We make three contributions. First, we formalize deductive, inductive, and abductive reasoning within the LLM context, connecting these classical cognitive paradigms with their algorithmic realizations. Second, we formalize adaptive reasoning as a control-augmented policy optimization problem balancing task performance with computational cost, distinguishing learned policies from inference-time control mechanisms. Third, we propose a systematic taxonomy organizing existing methods into training-based approaches that internalize adaptivity through reinforcement learning, supervised fine-tuning, and learned controllers, and training-free approaches that achieve adaptivity through prompt conditioning, feedback-driven halting, and modular composition. This framework clarifies how different mechanisms realize adaptive reasoning in practice and enables systematic comparison across diverse strategies. We conclude by identifying open challenges in self-evaluation, meta-reasoning, and human-aligned reasoning control.