Adnan KhanAdnanProduct & AI
San Francisco Bay Area
Get in touch

Source note3 min read

AREX: Towards a Recursively Self-Improving Agent for Deep Research

Source

arXiv

arxiv.org/abs/2607.21461v1

Read the source

Source

  • Source name: arXiv
  • Source kind: paper
  • Authors / organization: Shuqi Lu, Chaofan Li, Kun Luo, Zhang Zhang, Hui Wang, Hongwang Xiao, Zheng Liu, Lei Xiong, Jiahao Wang, Sen Wang, Xiyan Jiang, Wanli Li, Yuyang Hu, Hongjin Qian, Bingyu Yan, Ziyi Xia, Yingxia Shao, Kang Liu, Zhicheng Dou, Di He, Chaozhuo Li, Qiwei Ye, Zhongyuan Wang, Zheng Liu
  • Venue: arXiv: cs.AI
  • Published: 2026-07-23
  • Collected: 2026-07-24
  • URL: http://arxiv.org/abs/2607.21461v1
  • PDF: https://arxiv.org/pdf/2607.21461v1
  • DOI: UNKNOWN

Plain-Language Brief

The paper introduces a new type of research agent called AREX, which can recursively improve its answers by verifying intermediate results and using the partially verified state to guide subsequent refinement. The agent learns an autonomous context-update tool that compresses interaction history into a compact improvement state. It is trained on verified synthetic tasks and high-quality trajectories through agentic mid-training and long-horizon reinforcement learning. The model outperforms comparable-scale baselines across various benchmarks.

What the Document Says

  • Deep research requires finding answers that satisfy multiple constraints, but verification is often easier than discovery.
  • AREX is a family of Recursively Self-Improving (RSI) deep research agents that alternate between an inner research loop and an outer self-improvement loop.
  • The inner research loop gathers evidence and constructs a provisional answer, while the outer loop audits the answer constraint-wise, identifies unresolved claims, and launches targeted follow-up research.
  • AREX learns an autonomous context-update tool that compresses interaction history into a compact improvement state.
  • The paper mentions training AREX on verified synthetic tasks and high-quality trajectories through agentic mid-training and long-horizon reinforcement learning.
  • The model is tested across BrowseComp, WideSearch, DeepSearchQA, Humanity's Last Exam (HLE), and other reasoning and tool-use benchmarks.
  • AREX outperforms comparable-scale baselines and remains competitive with models using substantially more activated parameters.

Key Concepts Mentioned

  • deep research
  • Recursively Self-Improving (RSI) agents
  • inner research loop
  • outer self-improvement loop
  • context-update tool
  • agentic mid-training
  • long-horizon reinforcement learning
  • verified synthetic tasks
  • high-quality trajectories
  • BrowseComp
  • WideSearch
  • DeepSearchQA
  • Humanity's Last Exam (HLE)

Claims Made by the Document

  • AREX outperforms comparable-scale baselines and remains competitive with models using substantially more activated parameters.

Evidence Used

  • The model outperforms comparable-scale baselines across various benchmarks as stated in the text.

Methods / Approach

  • agentic mid-training
  • long-horizon reinforcement learning

Limitations Stated by the Authors

  • UNKNOWN
  • The paper does not explicitly state any limitations of their method according to the provided text.

Limitations Noted by the Agent

Agent interpretation — not from the source.

  • The model's performance may depend on the quality and relevance of the data it is trained on.
  • The model may have difficulties in generalizing to tasks outside its training domain.
  • The model's performance might be influenced by biases present in the training data.

Open Questions

  • What specific techniques does AREX use for targeted follow-up research?
  • How does AREX learn and update its context-update tool?

Notes for Future Q&A

May include interpretation. Labeled where it does.

  • The paper does not provide detailed information on how the targeted follow-up research is conducted.
  • The process of learning and updating the context-update tool is not described in detail.

Raw Abstract / Excerpt

Deep research requires agents to find answers that jointly satisfy multiple constraints. Discovering such answers is costly, whereas verifying a candidate can often be decomposed into tractable constraint-wise checks. This discovery--verification asymmetry suggests that a research agent should do more than simply search longer: it should recursively improve its current answer by verifying intermediate results and using the partially verified state to guide subsequent refinement. We introduce AREX, a family of Recursively Self-Improving (RSI) deep research agents. AREX alternates between an inner research loop that gathers evidence and constructs a provisional answer, and an outer self-improvement loop that audits the answer constraint-wise, identifies unresolved claims, and launches targeted follow-up research. To sustain RSI over long horizons, AREX learns an autonomous context-update tool that compresses growing interaction history into a compact improvement state preserving verified evidence and unresolved constraints, without relying on an external model. We train AREX on verified synthetic tasks and high-quality trajectories through agentic mid-training and long-horizon reinforcement learning. To mitigate sparse final rewards during long horizon learning, we emphasize key steps where decisive evidence is acquired or erroneous research directions are corrected. We instantiate a dense 4B model and a 122B-A10B Mixture-of-Experts model. Across BrowseComp, WideSearch, DeepSearchQA, Humanity's Last Exam (HLE), and other reasoning and tool-use benchmarks, AREX substantially outperforms comparable-scale baselines and remains competitive with models using substantially more activated parameters.