Fluid Representations in Reasoning Models

TL;DR

This paper investigates how reasoning language models develop internal structural representations during problem solving, revealing that in-context refinement of token representations enhances reasoning capabilities.

Contribution

It provides a mechanistic analysis of how a reasoning model refines internal representations, introducing the concept of Fluid Reasoning Representations and demonstrating their causal role in reasoning.

Findings

Models develop abstract, structure-focused encodings during reasoning.

Refined internal representations improve problem-solving accuracy.

In-context representation refinement is key to reasoning performance.

Abstract

Reasoning language models, which generate long chains of thought, dramatically outperform non-reasoning language models on abstract problems. However, the internal model mechanisms that allow this superior performance remain poorly understood. We present a mechanistic analysis of how QwQ-32B - a model specifically trained to produce extensive reasoning traces - process abstract structural information. On Mystery Blocksworld - a semantically obfuscated planning domain - we find that QwQ-32B gradually improves its internal representation of actions and concepts during reasoning. The model develops abstract encodings that focus on structure rather than specific action names. Through steering experiments, we establish causal evidence that these adaptations improve problem solving: injecting refined representations from successful traces boosts accuracy, while symbolic representations can replace many obfuscated encodings with minimal performance loss. We find that one of the factors driving reasoning model performance is in-context refinement of token representations, which we dub Fluid Reasoning Representations.