Answer Convergence as a Signal for Early Stopping in Reasoning
This work addresses the problem of high inference costs for real-world applications by making reasoning more efficient, though it is incremental as it builds on existing chain-of-thought methods.
The paper tackles the problem of verbose and redundant reasoning in chain-of-thought prompting for large language models, finding that models converge to final answers after 60% of reasoning steps, and proposes inference-time strategies that reduce token usage by over 40% on NaturalQuestions with little or no accuracy drop.
Chain-of-thought (CoT) prompting enhances reasoning in large language models (LLMs) but often leads to verbose and redundant outputs, thus increasing inference cost. We hypothesize that many reasoning steps are unnecessary for producing correct answers. To investigate this, we start with a systematic study to examine what is the minimum reasoning required for a model to reach a stable decision. We find that on math reasoning tasks like math, models typically converge to their final answers after 60\% of the reasoning steps, suggesting substantial redundancy in the remaining content. Based on these insights, we propose three inference-time strategies to improve efficiency: (1) early stopping via answer consistency, (2) boosting the probability of generating end-of-reasoning signals, and (3) a supervised method that learns when to stop based on internal activations. Experiments across five benchmarks and five open-weights LLMs show that our methods significantly reduce token usage with little or no accuracy drop. In particular, on NaturalQuestions, Answer Consistency reduces tokens by over 40\% while further improving accuracy. Our work underscores the importance of cost-effective reasoning methods that operate at inference time, offering practical benefits for real-world applications.