Metraipolis › archive › dc_5acdb076c46a
Iris: Climbing to the Search Frontier — arXiv:2609.04304
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.04304
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-16T22:18:24.359Z
- SHA-256 of the archived text
f54b39ee694c30e85ea2a096873f556976045184b9270d98846ae84bcf0ccc46- Stable URL of this page
https://metraipolis.com/documento/dc_5acdb076c46a
How to cite
Iris: Climbing to the Search Frontier — arXiv:2609.04304 — arXiv (en), https://arxiv.org/abs/2609.04304 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_5acdb076c46a (Metraipolis archive dc_5acdb076c46a, sha256 f54b39ee694c30e85ea2a096873f556976045184b9270d98846ae84bcf0ccc46)
BibTeX
@misc{dc5acdb076c46a,
title = {Iris: Climbing to the Search Frontier — arXiv:2609.04304},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.04304},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_5acdb076c46a; sha256 f54b39ee694c30e85ea2a096873f556976045184b9270d98846ae84bcf0ccc46},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.04304. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
We present Iris-mini and Iris-pro, two search agents trained at the 35B-A3B and 397B-A17B scales, together with the data pipeline and training recipe behind them. Tasks are reverse-constructed from the hyperlink structure of a web corpus: we author multi-hop chains over an entity graph distilled from a seed page and its out-links, rewrite every non-answer entity into a descriptive reference so that no clue can be resolved by string matching, and admit only questions that a reference model fails closed-book yet solves once the supporting evidence is supplied. These questions are then turned into trajectories, which are filtered at both the trajectory and the turn level before SFT. The policy is then optimized by RL against live search, with the reward judge and the observation summarizer served inside the training cluster, and with over-long rollouts interrupted at the request level and resumed from their committed prefix at the next step. We alternate the two stages in a procedure we call SFT-RL climbing, returning the hardest solved and most efficient rollouts of each RL round to the next supervised pass. Because inference-time context management is worth more on these benchmarks than most reported differences between systems, we evaluate every benchmark both with and without it, holding the tool set, the context limit, and the judge fixed. All results come from a single ReAct agent, with no sub-agents and no test-time verification. With management enabled, on BrowseComp, BrowseComp-ZH, DeepSearchQA, and HLE the two models reach $82.2/84.8/86.9/52.3$ and $88.6/85.1/92.9/56.4$, the strongest overall results among open-source search agents in their respective parameter ranges. We plan to release the model weights together with the complete recipe for data construction, training, and evaluation. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-07 in cs.AI). Categorie arXiv: Artificial Intelligence (cs.AI). Pagina del lavoro: https://arxiv.org/abs/2609.04304 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito). Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina. Letto il 2026-10-16T22:18:11.341Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-07?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).