PulseAugur
实时 23:26:42
English(EN) HARNESS-LM: A Three-Phase Training Recipe for Harnessing SLMs in Sponsored Search Retrieval

HARNESS-LM 训练紧凑型模型以实现更快的赞助搜索检索

研究人员开发了 HARNESS-LM (HLM),一种新颖的三阶段训练框架,旨在将大型语言模型的能力转移到紧凑、高效的模型中,用于赞助搜索检索。该方法包括训练一个高性能的“教师”模型,将其知识蒸馏到一个更小的“学生”编码器中,然后优化学生模型以获得最佳检索性能。HLM 成功恢复了教师模型超过 98% 的精度,同时显著降低了延迟并提高了吞吐量,通过在 Bing Ads 上的 A/B 测试证明了其实际效果。 AI

影响 使得强大的语言模型能够在对延迟敏感的应用中部署,从而提高赞助搜索等领域的效率和性能。

排序理由 发表了一篇详细介绍语言模型新训练框架的学术论文。

在 arXiv cs.IR (Information Retrieval) 阅读 →

AI 生成摘要 · Google Gemini · 来自 2 个来源。 我们如何撰写摘要 →

报道来源 [2]

  1. arXiv cs.AI TIER_1 English(EN) · Vipul Gupta, Shikhar Mohan, Lakshya Kumar, Pranjal Chitale, Nikit Begwani, Amit Singh, Manik Varma ·

    HARNESS-LM: A Three-Phase Training Recipe for Harnessing SLMs in Sponsored Search Retrieval

    arXiv:2605.23572v1 Announce Type: cross Abstract: In the competitive landscape of sponsored search, balancing retrieval quality with production latency is a critical challenge. While large retrieval models based on Small Language Models (SLMs) such as Qwen3-Embedding-4B/8B set st…

  2. arXiv cs.IR (Information Retrieval) TIER_1 English(EN) · Manik Varma ·

    HARNESS-LM: A Three-Phase Training Recipe for Harnessing SLMs in Sponsored Search Retrieval

    In the competitive landscape of sponsored search, balancing retrieval quality with production latency is a critical challenge. While large retrieval models based on Small Language Models (SLMs) such as Qwen3-Embedding-4B/8B set strong upper bounds on public benchmarks, their depl…