大语言模型何时应进行搜索？搜索路由的反事实监督
When Should LLMs Search? Counterfactual Supervision for Search Routing
AI总结：
研究大语言模型搜索路由问题，通过比较无搜索和强制搜索结果构建预言机，以其为评估标准和学习信号，用监督微调及偏好优化训练策略，提升了Gemma E2B和Qwen3.5 - 4B的路由宏F1，减少特定模型路由失败并揭示相关瓶颈。
摘要：
搜索增强语言模型可利用外部证据弥补参数知识的局限，但搜索并非总是有益：模型可能对已能回答的问题进行搜索，或在更适合纠正、澄清或弃权时依赖有噪声的证据。我们将此表述为实例级搜索路由问题：决定相对于不进行搜索的执行方式，是否需要搜索以提高任务成功率。为得出监督信息，我们比较同一问题的无搜索和强制搜索结果，并基于特定任务的成功率构建一个关于无搜索、搜索和未解决情况的预言机。将此预言机用作评估标准和学习信号，我们通过监督微调及偏好优化来训练搜索路由策略，提高了Gemma E2B在符合预言机条件的示例上的路由宏F1，从0.7082提升到0.8235，Qwen3.5 - 4B从0.7053提升到0.8365。进一步分析表明，学习到的策略减少了特定模型的路由失败：Gemma主要学会了不进行搜索的克制，而Qwen进一步减少了搜索遗漏；剩余的未解决情况揭示了涉及模型能力、检索预算、证据使用和策略行为的异质瓶颈。
英文摘要：
Search-augmented language models can use external evidence to compensate for limitations in parametric knowledge, but search is not uniformly beneficial: models may call search for questions they can already answer, or rely on noisy evidence when correction, clarification, or abstention would be more appropriate. We formulate this as an instance-level search-routing problem: deciding whether search is needed to improve task success relative to a no-search execution. To derive supervision, we compare no-search and forced-search outcomes for the same question and construct an oracle over NO SEARCH, SEARCH, and UNSOLVED based on task-specific success. Using this oracle as both an evaluation criterion and a learning signal, we train search-routing policies with supervised fine-tuning and preference optimization, improving routing macro-F1 on oracle-eligible examples from 0.7082 to 0.8235 for Gemma E2B and from 0.7053 to 0.8365 for Qwen3.5-4B. Further analysis shows that the learned policies reduce model-specific routing failures: Gemma primarily learns no-search restraint, while Qwen further reduces missed search; residual UNSOLVED cases reveal heterogeneous bottlenecks involving model capacity, retrieval budget, evidence use, and policy behavior.
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。
