不训练模型就能自动判断何时停止检索，节省调用次数同时保持高准确率。
TASR: Training-Free Adaptive Stopping for Iterative Retrieval
- 基于答案重复和校准后逻辑值差判断是否停止，无需训练任何组件。
- 在32种配置中保持94.8%的准确率，调用次数减少至固定k=5的62.6%。
- 适用于多种模型、检索器和数据集，适合追求高效可靠的生产系统。
迭代检索增强生成代理常因过早停止或过度检索而浪费资源，当前方法需训练停机策略，难以迁移。本文提出TASR（无需训练的自适应停机规则），仅通过一条判断条件：当模型重复上一轮归一化答案且等距校准后的逻辑值差超过0.25时触发停止。该规则不依赖分类器或价值头，阈值在32种（模型、检索器、语料）配置中固定不变。在3模型×2数据集的干扰项网格中，TASR以62.6%的调用次数保留了固定k=5的94.8%宏平均F1，优于固定k=3的+3.42 F1。在九个开放域BM25单元中实现55.01 F1（2.98次调用）优于固定k=3的54.33 F1（3.00次调用），并在两个检索器家族的密集检索单元及八组Nemotron-3-Ultra-550B生产模型中均无显著退化。该规则从381个候选规则中筛选得出，无其他方案在帕累托意义下更优。信号质量分析显示，经RLHF微调模型的置信度评分严重坍缩（96.5%为5，熵0.182纳特），而逻辑值差可实现40倍更好的类别条件分离，验证设计合理性。TASR是可审计、免训练的帕累托基准，代码开源于https://github.com/JSBAICenter/TASR。
原文摘要 · Abstract (English)
Iterative retrieval-augmented generation agents commonly overspend by continuing to retrieve after the model has converged on an answer, incurring calls that change neither the prediction nor the supporting evidence. Existing remedies learn a stopping policy from labeled trajectories, tying the decision to a trained component that requires retraining for each new model or task. We propose TASR (Training-Free Adaptive Stopping Rule), a one-line predicate that fires when the model repeats its previous-round normalized answer and the isotonically calibrated logit margin exceeds 0.25. No classifier or value head is learned; the threshold is fixed across all thirty-two (model, retriever, corpus) configurations we evaluate. On a 3-model x 2-dataset distractor grid, TASR retains 94.8% of fixed-k=5's macro F1 at 62.6% of its calls and exceeds fixed-k=3 by +3.42 F1. The pattern holds on nine open-domain BM25 cells (55.01 F1 at 2.98 calls vs. 54.33 at 3.00 for fixed-k=3) and, with calibration locked from the distractor split, on nine dense-retrieval cells across two retriever families, and on eight cells of a Nemotron-3-Ultra-550B production model, with zero significant regressions in any extension. The rule was selected from an exhaustive enumeration of 381 candidate stopping rules on the canonical selection cell, where no alternative Pareto-dominates it. A signal-quality analysis shows that verbalized 1-5 confidence collapses on RLHF-tuned models (96.5% of values equal 5, entropy 0.182 nats), while the logit margin achieves 40x better class-conditional separation, grounding the design in a measurable model pathology. TASR is an auditable, training-free Pareto baseline for adaptive stopping in iterative retrieval. Code is publicly available at https://github.com/JSBAICenter/TASR
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。