Reasoning Shift: How Context Silently Shortens LLM Reasoning

TL;DR

This paper investigates how different contexts influence the reasoning traces of large language models, revealing that context often causes models to produce shorter, less verified reasoning processes without affecting simple task performance.

Contribution

It systematically evaluates reasoning models across various scenarios, uncovering the phenomenon of context-induced reasoning trace compression and its implications.

Findings

Models produce up to 50% shorter reasoning traces in different contexts.

Shorter reasoning traces are linked to reduced self-verification behaviors.

Performance on simple problems remains unaffected despite reasoning trace compression.

Abstract

Large language models (LLMs) exhibiting test-time scaling behavior, such as extended reasoning traces and self-verification, have demonstrated remarkable performance on complex, long-term reasoning tasks. However, the robustness of these reasoning behaviors remains underexplored. To investigate this, we conduct a systematic evaluation of multiple reasoning models across three scenarios: (1) problems augmented with lengthy, irrelevant context; (2) multi-turn conversational settings with independent tasks; and (3) problems presented as a subtask within a complex task. We observe an interesting phenomenon: reasoning models tend to produce much shorter reasoning traces (up to 50%) for the same problem under different context conditions compared to the traces produced when the problem is presented in isolation. A finer-grained analysis reveals that this compression is associated with a decrease in self-verification and uncertainty management behaviors, such as double-checking. While this behavioral shift does not compromise performance on straightforward problems, it might affect performance on more challenging tasks. We hope our findings draw additional attention to both the robustness of reasoning models and the problem of context management for LLMs and LLM-based agents.