Paper Detail

T-Search: An Open Agentic Retriever and Playground for Hard Multi-Step Search

Olga Tsymboi, Ramil Latypov, Aleksandr Medvedev, Danil Taranets, Dmitrii Stoianov, Nikita Gulyakov, Gleb Alektorov, Anatolii Potapov

arxiv Score 7.8

Published 2026-10-05 · First seen 2026-10-06

General AI

Abstract

We present T-Search, an open-weight agentic retriever for hard multi-step search. Given a question and a search tool over a fixed corpus, it runs a bounded multi-round search and returns a ranked list of evidence chunks with short justifications, leaving answer generation to a downstream model, so backend and generator can be swapped without retraining. T-Search is built on Qwen3.6-35B-A3B and trained on adversarially filtered synthetic search tasks with round-sliced supervised fine-tuning followed by GSPO on a recall reward. Averaged over seven English and Russian benchmarks with gold evidence annotations, it reaches 56.0 Recall@10 with one rollout, 14.4 points above its base, and 61.3 with three fused rollouts, outperforming larger open models. We release the model, harness, live demo, and three benchmarks, including TRuST, the first native-Russian hard-search benchmark.

Workflow Status

Review status
pending
Role
unreviewed
Read priority
soon
Vote
Not set.
Saved
no
Collections
Not filed yet.
Next action
Not filled yet.

Reading Brief

No structured notes yet. Add `summary_sections`, `why_relevant`, `claim_impact`, or `next_action` in `papers.jsonl` to enrich this view.

Why It Surfaced

No ranking explanation is available yet.

Tags

No tags.

BibTeX

@article{tsymboi2026t,
  title = {T-Search: An Open Agentic Retriever and Playground for Hard Multi-Step Search},
  author = {Olga Tsymboi and Ramil Latypov and Aleksandr Medvedev and Danil Taranets and Dmitrii Stoianov and Nikita Gulyakov and Gleb Alektorov and Anatolii Potapov},
  year = {2026},
  abstract = {We present T-Search, an open-weight agentic retriever for hard multi-step search. Given a question and a search tool over a fixed corpus, it runs a bounded multi-round search and returns a ranked list of evidence chunks with short justifications, leaving answer generation to a downstream model, so backend and generator can be swapped without retraining. T-Search is built on Qwen3.6-35B-A3B and trained on adversarially filtered synthetic search tasks with round-sliced supervised fine-tuning follo},
  url = {https://arxiv.org/abs/2610.06782},
  keywords = {cs.CL},
  eprint = {2610.06782},
  archiveprefix = {arXiv},
}

Metadata

{}