How Is LLM Reasoning Distracted by Irrelevant Context? An Analysis Using a Controlled Benchmark
This addresses the robustness of LLMs in reasoning tasks for AI researchers, though it is incremental as it builds on existing benchmarks and methods.
The authors tackled the problem of LLMs being distracted by irrelevant context in reasoning tasks, showing that LLMs are significantly sensitive to such distractions, affecting reasoning path selection and arithmetic accuracy, and that training with strong distractors improves performance in both in-distribution and out-of-distribution scenarios.
We introduce Grade School Math with Distracting Context (GSM-DC), a synthetic benchmark to evaluate Large Language Models' (LLMs) reasoning robustness against systematically controlled irrelevant context (IC). GSM-DC constructs symbolic reasoning graphs with precise distractor injections, enabling rigorous, reproducible evaluation. Our experiments demonstrate that LLMs are significantly sensitive to IC, affecting both reasoning path selection and arithmetic accuracy. Additionally, training models with strong distractors improves performance in both in-distribution and out-of-distribution scenarios. We further propose a stepwise tree search guided by a process reward model, which notably enhances robustness in out-of-distribution conditions.