In deep research, agents need to find answers that satisfy multiple constraints. The discovery process is costly, while verifying a candidate can often be decomposed into tractable constraint-wise checks. This discovery-verification asymmetry suggests that a research agent should not only search longer but also recursively improve its current answer by verifying intermediate results and using the partially verified state to guide subsequent refinements.
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.
Blogger's Review: AREX significantly enhances the efficiency of deep research through its recursive self-improvement mechanism, particularly excelling in handling complex constraints. Its innovative context-update tool effectively supports long-horizon learning, showcasing the future direction for research agents. The model's outstanding performance across various benchmarks indicates its broad application potential in the field of deep learning.