ConfSpec：通过置信度门控验证实现高效的步骤级推测推理
ConfSpec: Efficient Step-Level Speculative Reasoning via Confidence-Gated Verification
AI总结：
ConfSpec通过置信度门控验证框架，在保持准确性的同时，实现高效步骤级推测推理，达到2.24倍的端到端加速。
摘要：
链式思维推理显著提高了大型语言模型在复杂任务上的性能，但由于生成长轨迹导致推理延迟较高。步骤级推测推理旨在缓解这一成本，但现有方法面临准确率、推理速度和资源效率之间的长期权衡。我们提出了ConfSpec，一种置信度门控级联验证框架，解决了这一权衡。我们的关键见解是生成与验证之间的不对称性：生成正确的推理步骤需要大量模型容量，而步骤级验证是一个受约束的判别任务，小型草稿模型在其能力范围内校准良好，使高置信度的草稿决策可以直接接受，同时选择性地将不确定的案例提升到大型目标模型。在多样化的负载评估中，ConfSpec实现了高达2.24倍的端到端加速，同时保持目标模型的准确性。我们的方法不需要外部判断模型，并且与令牌级推测解码正交，能够进一步乘法加速。
英文摘要：
Chain-of-Thought reasoning significantly improves the performance of large language models on complex tasks, but incurs high inference latency due to long generation traces. Step-level speculative reasoning aims to mitigate this cost, yet existing approaches face a long-standing trade-off among accuracy, inference speed, and resource efficiency. We propose ConfSpec, a confidence-gated cascaded verification framework that resolves this trade-off. Our key insight is an asymmetry between generation and verification: while generating a correct reasoning step requires substantial model capacity, step-level verification is a constrained discriminative task for which small draft models are well-calibrated within their competence range, enabling high-confidence draft decisions to be accepted directly while selectively escalating uncertain cases to the large target model. Evaluation across diverse workloads shows that ConfSpec achieves up to 2.24$\times$ end-to-end speedups while matching target-model accuracy. Our method requires no external judge models and is orthogonal to token-level speculative decoding, enabling further multiplicative acceleration.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
