超越并行采样：代理搜索的多样化查询初始化

arXiv:2606.17209v1 公告类型：新摘要：代理搜索的测试时间扩展通常会增加深度（即每个轨迹更多的转弯和标记）或广度（即更多的并行推出）。在这里，我们关注广度缩放，表明标准并行采样产生的回报递减，将其追溯到第一轮的查询冗余。当模型在各个部署中发出类似的第一个查询时，线程会检索重叠的证据，并且后续轮次以该共享检索为条件。 We address this limitation with DivInit, a training-free intervention at the first turn. Rather than sampling k independent first queries, DivInit draws n candidates from a single call, picks k < n diverse seeds, and runs them as parallel trajectories. Across five open-weight models and eight benchmarks, DivInit consistently improves over standard parallel sampling, with average gains of five to seven points on multi-hop QA at matched compute. Code available at https://github.com/cxcscmu/diverse-query-initialization