Joint or Per-Option? Compact Encoders for Closed-Set Decisions from 17 Million to 1 Billion Parameters
作者: Tardelli Ronan Coelho Stekel
分类: cs.CL, cs.LG
发布日期: 2026-10-22
备注: 18 pages, 8 figures, 8 tables. Code: https://github.com/tardellirs/crivo-decision-encoders ; models: https://huggingface.co/tardellirs
💡 一句话要点
提出紧凑编码器以优化闭集决策过程
🎯 匹配领域: 支柱九:具身大模型 (Embodied Foundation Models)
关键词: 闭集决策 紧凑编码器 每选项评分 联合评分 语言模型 性能优化 智能助手
📋 核心要点
- 现有方法在闭集决策中使用大型语言模型时,速度慢且对输入顺序敏感,影响决策效率。
- 论文提出了一种紧凑的编码器设计,探索每选项评分与联合评分的性能差异,以优化决策过程。
- 实验结果表明,32百万参数的每选项评分在多个基准任务上显著优于联合评分,且在处理速度上表现更佳。
📝 摘要(中文)
许多语言模型管道以闭集决策结束,例如选择工具。使用大型语言模型进行提示既缓慢又对顺序敏感。本文探讨了编码器的最小化及其应如何读取选项。通过三种种子研究,发现每选项评分在32百万参数下优于联合评分,且在多个基准任务中表现出显著提升。六个发布的联合编码器在准确性上超过了435百万参数的DeBERTa-v3零-shot分类器,且在处理速度上具有明显优势。
🔬 方法详解
问题定义:本文旨在解决在闭集决策中使用大型语言模型的速度和顺序敏感性问题,现有方法在处理效率上存在不足。
核心思路:论文提出通过紧凑编码器设计,采用每选项评分方式来提高决策效率,避免联合评分的延迟和不确定性。
技术框架:整体架构包括编码器模块、评分机制和重排序初始化。编码器负责处理输入选项,评分机制根据每个选项的表现进行评估,重排序初始化则确保输入与每选项读取相匹配。
关键创新:最重要的创新在于提出了每选项评分的策略,显著提高了决策的准确性和效率,尤其是在参数较少的情况下。与现有方法相比,能够在相同的计算资源下实现更好的性能。
关键设计:在参数设置上,使用了32百万至1亿参数的编码器,损失函数设计为优化每选项的评分,网络结构采用了适应性重排序机制,以确保在不同初始化条件下的表现一致性。
📊 实验亮点
实验结果显示,32百万参数的每选项评分在37个基准任务上比联合评分提高了5.8个点的准确性,且在处理速度上,最小模型在16个处理线程上仅需14毫秒,远低于传统分类器的728毫秒。多个基准任务验证了该方法的有效性。
🎯 应用场景
该研究的潜在应用场景包括智能助手、自动化工具选择和决策支持系统等领域。通过优化闭集决策过程,可以显著提高系统的响应速度和准确性,进而提升用户体验和工作效率。未来,该技术有望在更多实时决策场景中得到应用。
📄 摘要(原文)
Many language-model pipelines end in a closed-set decision, such as selecting a tool. Prompting a large language model for it is slow and order-sensitive. We ask how small an encoder for such decisions can be and whether it should read the options jointly or separately. In a three-seed study at 32 million parameters, per-option scoring beats joint scoring on a held-out sample of a 37-benchmark decision suite by 5.8 points of chance-corrected skill at matched optimizer steps and by 2.8 at approximately matched training compute; two external benchmarks agree, though not significantly on their unseen tasks. Crossing reader with initialisation shows that the advantage requires a reranker initialisation, whose query-candidate pair input matches per-option reading; from the raw encoder the two readers are indistinguishable. Per-option scoring needs 5 to 13 times more encoder passes: at most twice the median latency, but more than four times the mean. Of six released joint encoders (17 million to 1 billion parameters), the 68- and 150-million-parameter models exceed a 435-million-parameter DeBERTa-v3 zero-shot classifier by 4.9 and 5.9 accuracy points on 209 overlap-audited System1-Arena tasks and are at least on par with it on the 145 tasks unseen in training. The smallest answers a median request in 14 milliseconds on 16 processor threads, against 728 for the classifier. Reordering options changes 12 to 23% of the joint models' answers (per-option scoring: none). Up to 68 million parameters, they lead their size range on a public leaderboard's private, mainly in-distribution tests.