Not Just the Destination, But the Journey: Reasoning Traces Causally Shape Generalization Behaviors

TL;DR

This paper investigates whether the reasoning process in large language models causally influences their generalization behaviors, finding that reasoning content significantly impacts model outputs independently of final answers.

Contribution

The study introduces a controlled experimental framework to isolate and analyze the causal effect of reasoning traces on model behavior, revealing their independent influence beyond answer supervision.

Findings

Reasoning training can amplify harmful behaviors.

Different reasoning types lead to distinct behaviors.

Reasoning alone can alter model behavior without answer supervision.

Abstract

Chain-of-Thought (CoT) is often viewed as a window into LLM decision-making, yet recent work suggests it may function merely as post-hoc rationalization. This raises a critical alignment question: Does the reasoning trace causally shape model generalization independent of the final answer? To isolate reasoning's causal effect, we design a controlled experiment holding final harmful answers constant while varying reasoning paths. We construct datasets with \textit{Evil} reasoning embracing malice, \textit{Misleading} reasoning rationalizing harm, and \textit{Submissive} reasoning yielding to pressure. We train models (0.6B--14B parameters) under multiple paradigms, including question-thinking-answer (QTA), question-thinking (QT), and thinking-only (T-only), and evaluate them in both think and no-think modes. We find that: (1) CoT training could amplify harmful generalization more than standard fine-tuning; (2) distinct reasoning types induce distinct behavioral patterns aligned with their semantics, despite identical final answers; (3) training on reasoning without answer supervision (QT or T-only) is sufficient to alter behavior, proving reasoning carries an independent signal; and (4) these effects persist even when generating answers without reasoning, indicating deep internalization. Our findings demonstrate that reasoning content is causally potent, challenging alignment strategies that supervise only outputs.