用重置丢弃法提升大模型推理效率，相同预算下答更多题。
More Bang for the Buck: Improving the Inference of Large Language Models at a Fixed Budget using Reset and Discard (ReD)
- 提出重置丢弃策略，避免重复尝试同一问题
- 相同预算下覆盖题目数提升30%以上，节省成本
- 适合需要高效推理的开发者和研究者
大语言模型在可验证任务上的性能通常以pass@k衡量，即k次尝试中至少一次答对的概率。但在固定预算下，更合适的指标是coverage@cost，即总尝试次数下的唯一正确问题数。本文揭示pass@k的幂律特性导致coverage@cost呈亚线性增长（收益递减）。为此提出重置与丢弃（ReD）方法，无论pass@k形式如何，均能提升给定预算下的覆盖率。若已知pass@k，可定量预测尝试次数节省；若未知，ReD可推断其幂律指数。在HumanEval、GSM8K和MMLU-Pro三个基准上的实验表明，ReD显著降低达到目标覆盖率所需的尝试次数、生成token数及美元成本，且对不完美验证器鲁棒，优于现有分配基线。
原文摘要 · Abstract (English)
The performance of large language models (LLMs) on verifiable tasks is usually measured by pass@k, the probability of answering a question correctly at least once in k trials. At a fixed budget, a more suitable metric is coverage@cost, the average number of unique questions answered as a function of the total number of attempts. We connect the two metrics and show that the empirically-observed power-law behavior in pass@k leads to a sublinear growth of the coverage@cost (diminishing returns). To solve this problem, we propose Reset-and-Discard (ReD), a query method of LLMs that increases coverage@cost for a given budget, regardless of the pass@k form. Moreover, given a pass@k, we can quantitatively predict the savings in the total number of attempts using ReD. If pass@k is not available for the model, ReD can infer its power-law exponent. Experiments on three LLMs across coding (HumanEval), math (GSM8K), and reasoning (MMLU-Pro) benchmarks demonstrate that ReD substantially reduces the required attempts, tokens, and USD cost to reach a desired coverage, while also offering an efficient way to measure inference power-laws. ReD's advantage is maintained for imperfect verifiers and outperforms the tested allocation baselines.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。