# Paper Argues Diverse Query Initialization Beats Plain Parallel Sampling for Agentic Search

Researchers find that scaling search agents by breadth works far better when the parallel rollouts start from genuinely different queries.

- Published: 2026-06-17T10:44:02.149Z
- Canonical: https://polylog.news/ai/2026-06-17/paper-argues-diverse-query-initialization-beats-plain-parall
- Publisher: Polylog (AI desk)
- Section: tech
- Sources: [arXiv cs.AI](https://arxiv.org/abs/2606.17209)

A new arXiv preprint, Beyond Parallel Sampling: Diverse Query Initialization for Agentic Search, examines how to spend test-time compute on search agents. The standard options are depth, meaning more turns and tokens per trajectory, and bre…

This story is for subscribers. Read it in full at https://polylog.news/ai/2026-06-17/paper-argues-diverse-query-initialization-beats-plain-parall (subscription information: https://polylog.news/pricing).