Skip to content
AI News HubLIVE
Original source2 min read

Iris: Climbing to the Search Frontier

Summary

The paper introduces Iris-mini and Iris-pro, open-source search agents trained at 35B-A3B and 397B-A17B scales. Tasks are reverse-constructed from web hyperlink structures to create multi-hop questions, then filtered for SFT and optimized via RL against live search. With inference-time context management enabled, the models achieve strong results across BrowseComp, BrowseComp-ZH, DeepSearchQA, and HLE.

SourcearXiv AIAuthor: Ziyuan Liu, Hengqi Liu, Zichuan Wang, Yang Qin, Jiachen Liang, Xu Chu, Shaowei Chen, Yuantao Gu, Mu Chuan
Iris: Climbing to the Search Frontier
Report an error

The correction channel is not available yet. You can copy the article reference below for later.

Correction instructions
Read article

[Submitted on 3 Sep 2026]

Title:Iris: Climbing to the Search Frontier

View a PDF of the paper titled Iris: Climbing to the Search Frontier, by Ziyuan Liu and 8 other authors

View PDF HTML (experimental)

Abstract: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.

Comments: 12 pages, 2 figures

Subjects:

Artificial Intelligence (cs.AI)

Cite as: arXiv:2609.04304 [cs.AI]

(or arXiv:2609.04304v1 [cs.AI] for this version)

https://doi.org/10.48550/arXiv.2609.04304

arXiv-issued DOI via DataCite (pending registration)

Submission history

From: Ziyuan Liu [view email] [v1] Thu, 3 Sep 2026 17:51:01 UTC (2,175 KB)

Full-text links:

Access Paper:

View a PDF of the paper titled Iris: Climbing to the Search Frontier, by Ziyuan Liu and 8 other authors

View PDF

HTML (experimental)

TeX Source

view license

Current browse context:

cs.AI

new | recent | 2026-09

Change to browse by:

cs

References & Citations

NASA ADS

Google Scholar

Semantic Scholar

Loading...

Data provided by:

Bibliographic Tools

Bibliographic and Citation Tools

Bibliographic Explorer Toggle

Bibliographic Explorer (What is the Explorer?)

Connected Papers Toggle

Connected Papers (What is Connected Papers?)

Litmaps Toggle

Litmaps (What is Litmaps?)

scite.ai Toggle

scite Smart Citations (What are Smart Citations?)

Code, Data, Media

Code, Data and Media Associated with this Article

alphaXiv Toggle

alphaXiv (What is alphaXiv?)

Links to Code Toggle

CatalyzeX Code Finder for Papers (What is CatalyzeX?)

DagsHub Toggle

DagsHub (What is DagsHub?)

GotitPub Toggle

Gotit.pub (What is GotitPub?)

Huggingface Toggle

Hugging Face (What is Huggingface?)

ScienceCast Toggle

ScienceCast (What is ScienceCast?)

Demos

Demos

Replicate Toggle

Replicate (What is Replicate?)

Spaces Toggle

Hugging Face Spaces (What is Spaces?)

Spaces Toggle

TXYZ.AI (What is TXYZ.AI?)

Related Papers

Recommenders and Search Tools

Link to Influence Flower

Influence Flower (What are Influence Flowers?)

Core recommender toggle

CORE Recommender (What is CORE?)

Author

Venue

Institution

Topic

About arXivLabs

arXivLabs: experimental projects with community collaborators

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.

Which authors of this paper are endorsers? | Disable MathJax (What is MathJax?)

Key points and analysis

Article intelligence

EngineersAdvanced

Key points

  • Iris-mini and Iris-pro are open-source search agents at two parameter scales, based on single ReAct agents.
  • Training data is synthesized from web link structures, ensuring multi-hop questions cannot be solved by closed-book knowledge or string matching.
  • SFT-RL climbing alternates supervised fine-tuning with reinforcement learning over live search.
  • Enabled context management yields the strongest open-source results on BrowseComp, DeepSearchQA, and HLE among comparable parameter ranges.

Highlights and analysis are generated automatically and may contain errors. Check the original source.