T-Search: open agentic retriever hits 61.3 Recall@10, beats larger open models
_reachsumit · x · 2026-10-06
T-Search is an open-weight agentic retriever for hard multi-step search: given a question and a search tool over a fixed corpus, it runs bounded multi-round search and returns ranked evidence chunks with justifications, leaving generation to a swappable downstream model. Built on Qwen3.6-35B-A3B, trained with round-sliced SFT then GSPO on a recall reward over adversarially filtered synthetic tasks. Averaged over 7 English and Russian benchmarks, it reaches 56.0 Recall@10 with one rollout (+14.4 over base) and 61.3 with three fused rollouts, outperforming larger open models. Model, harness, live demo and three benchmarks released, including TRuST, the first native-Russian hard-search benchmark.
More from coding & agent
- Matt Pocock: Customize Your Coding Agent Skills to Your Own Workflow, Don't Use Them Off-the-Shelf — mattpocockuk · 2026-10-06
- Stanford ACE team unveils Sentry: failure tips in context hurt LLM agents, +39% gains — StanfordAILab · 2026-10-06
- Shadowrocket TUN-Only Setup to Stop Claude Account Bans — aigclink · 2026-10-06
- OpenAI streamlines ChatGPT plugin submissions: upload zip, fix validation, publish — Dimillian · 2026-10-06
- One tool beats two: how combining fetch and extraction fixed my agent's context overflow — OkShirt9372 · 2026-10-06
- Long-running benchmarks find Strata inference server failing full-build scenarios — julianharris · 2026-10-06