---
title: "AREX：迈向递归自我改进的深度研究智能体"
source: "arXiv AI/CL/LG（论文）"
source_tier: "t2"
url: "https://arxiv.org/abs/2607.21461v1"
permalink: "https://feed.astralor.com/items/itm_2af7e9352c3787b41a0661d5"
published: "2026-07-23T16:05:46.000Z"
score: 81
channel: "ai"
tags: ["AREX","arXiv","深度研究智能体","混合专家模型","递归自我改进"]
---

# AREX：迈向递归自我改进的深度研究智能体

## AI 点评

> 这篇论文展示了如何通过递归自我改进突破深度研究的瓶颈，对构建能自主完成复杂调研任务的AI智能体开发者尤为重要。它提出的上下文压缩工具解决了长程交互的核心难题，而开放模型和训练数据的做法将加速该领域进展。

## 速览

AREX是一种递归自我改进的深度研究智能体框架，通过交替进行内部研究循环和外部自我改进循环来逐步优化答案。内部循环收集证据并构建初步答案，外部循环按约束逐项审计，识别未解决的声明并启动针对性研究。AREX学习了一个自主上下文更新工具，将增长的历史交互压缩为紧凑的改进状态，无需外部模型。在4B和122B-A10B混合专家模型上训练，在BrowseComp、WideSearch、DeepSearchQA、人类最后考试等基准上显著优于同规模基线，与更大模型竞争。

## 关键点

- AREX提出递归自我改进（RSI）范式，通过内外双循环迭代优化答案
- 学习自主上下文更新工具，压缩历史交互为紧凑改进状态，消除外部模型依赖
- 在多个深度研究和推理基准上超越同规模基线，与更大激活参数模型竞争
- 支持从4B到122B-A10B混合专家模型的规模扩展

## 全文

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.

---

引用前请核对原文;本文由星际流动整理:https://feed.astralor.com/items/itm_2af7e9352c3787b41a0661d5