在大语言模型中的确定性引导推理：一种动态思维预算方法
Certainty-Guided Reasoning in Large Language Models: A Dynamic Thinking Budget Approach
- Nokia Bell Labs, France（诺基亚贝尔实验室，法国）
机构由 AI 辅助整理，请以论文原文为准。
AI总结：
CGR通过动态调整推理预算，在保持准确性的同时减少token使用，通过确定性评估实现高效的推理过程。
AI中文摘要：
大型推理语言模型通常使用固定的推理预算，这可能导致计算浪费或提前终止推理。我们引入了确定性引导推理（CGR），这是一种模型无关的自适应推理过程，定期探测当前推理是否支持有信心的最终答案，并在达到目标确定性阈值时提前终止，否则继续直到结束思考的标记或预算限制。确定性是从模型对答案标记的预测概率中估计的，从而得到一个轻量的停止标准。在AIME2025上，CGR保持基线准确性的同时减少token使用，提供了一个可调节的确定性-效率权衡，可以消除数百万个token。在64个随机种子中，CGR表现出一致的行为。我们还引入了一个Grade指标，惩罚错误答案并允许弃权，捕捉风险敏感的性能。结果表明，CGR在确定性较低时通过弃权来提高Grade。
英文摘要：
Large reasoning language models are typically run with fixed inference budgets, which can waste computation or terminate reasoning prematurely. We introduce Certainty-Guided Reasoning (CGR), a model-agnostic adaptive inference procedure that periodically probes whether the current reasoning supports a confident final answer and terminates early once a target certainty threshold is reached, otherwise continuing until the end-of-thinking token or the budget limit. Certainty is estimated from the model's predicted probabilities over the answer tokens, yielding a lightweight stopping criterion. On AIME2025, CGR preserves baseline accuracy while reducing token usage, providing a tunable certainty-efficiency trade-off that can eliminate millions of tokens in aggregate. Across 64 random seeds, CGR exhibits consistent behavior. We also introduce a Grade metric that penalizes incorrect answers and permits abstention, capturing risk-sensitive performance. Results show that CGR improves Grade by abstaining when certainty remains low.
