InjectRBP: Steering Large Language Model Reasoning Behavior via Pattern Injection

TL;DR

This paper introduces InjectRBP, a method to steer large language model reasoning by injecting behavioral patterns without parameter updates, leading to improved reasoning performance across tasks.

Contribution

It systematically analyzes reasoning behavioral patterns and proposes two parameter-free methods, InjectCorrect and InjectRLOpt, to enhance reasoning quality by pattern injection.

Findings

InjectCorrect improves reasoning accuracy by up to 5.34%.

InjectRLOpt achieves up to 8.67% performance gain.

Behavioral pattern injection significantly influences reasoning outcomes.

Abstract

Reasoning can significantly enhance the performance of Large Language Models. While recent studies have exploited behavior-related prompts adjustment to enhance reasoning, these designs remain largely intuitive and lack a systematic analysis of the underlying behavioral patterns. Motivated by this, we investigate how models' reasoning behaviors shape reasoning from the perspective of behavioral patterns. We observe that models exhibit adaptive distributions of reasoning behaviors when responding to specific types of questions, and that structurally injecting these patterns can substantially influence the quality of the models' reasoning processes and outcomes. Building on these findings, we propose two optimization methods that require no parameter updates: InjectCorrect and InjectRLOpt. InjectCorrect guides the model by imitating behavioral patterns derived from its own past correct answers. InjectRLOpt learns a value function from historical behavior-pattern data and, via our proposed Reliability-Aware Softmax Policy, generates behavioral injectant during inference to steer the reasoning process. Our experiments demonstrate that both methods can improve model performance across various reasoning tasks without requiring any modifications to model parameters, achieving gains of up to 5.34% and 8.67%, respectively.