Dynamic Thinking-Token Selection for Efficient Reasoning in Large Reasoning Models

TL;DR

This paper introduces DynTS, a method that improves the efficiency of large reasoning models by selectively retaining only decision-critical tokens' cache states during inference, reducing memory and computation costs.

Contribution

It proposes a novel token selection approach based on attention maps to identify and retain only decision-critical tokens, enhancing reasoning model efficiency.

Findings

Significantly reduces memory footprint during inference.

Maintains reasoning accuracy with selective token retention.

Demonstrates improved efficiency over baseline models.

Abstract

Large Reasoning Models (LRMs) excel at solving complex problems by explicitly generating a reasoning trace before deriving the final answer. However, these extended generations incur substantial memory footprint and computational overhead, bottlenecking LRMs' efficiency. This work uses attention maps to analyze the influence of reasoning traces and uncover an interesting phenomenon: only some decision-critical tokens in a reasoning trace steer the model toward the final answer, while the remaining tokens contribute negligibly. Building on this observation, we propose Dynamic Thinking-Token Selection (DynTS). This method identifies decision-critical tokens and retains only their associated Key-Value (KV) cache states during inference, evicting the remaining redundant entries to optimize efficiency.