AREX-2: Advancing Self-Improving Agents through Long-Horizon Reflective Tasks

AREX-2：通过长程反思任务推进自我改进智能体

Abstract: We present AREX-2, an effort to advance the self-improving capability of LLM agents, which we define as the ability to iteratively refine a solution at test time. This ability rests on two complementary capabilities: reflection, which produces a solution better than the current one, and long-horizon execution, which keeps the iteration effective over many rounds.

摘要： 我们提出了 AREX-2，旨在提升大语言模型（LLM）智能体的自我改进能力。我们将这种能力定义为在测试阶段迭代优化解决方案的能力。该能力依赖于两个互补的要素：一是“反思”，即产生优于当前方案的改进方案；二是“长程执行”，即确保迭代在多轮过程中持续有效。

We hypothesize that both capabilities are domain-agnostic, and can therefore be learned in scenarios that are well suited for supervision. Accordingly, we synthesize long-horizon improvement trajectories from machine learning and algorithmic programming tasks, two domains that offer verifiable feedback and reward sustained iteration.

我们假设这两种能力与领域无关，因此可以在适合监督学习的场景中进行训练。据此，我们从机器学习和算法编程任务中合成了长程改进轨迹；这两个领域能够提供可验证的反馈，并奖励持续的迭代过程。

Trained on this data, our agent, built on Qwen3.8-27B, achieves strong results on MLE-bench Lite (81.8) and Frontier-CS (70.7), transfers to deep research with 84.0 on BrowseComp, 52.6 on HLE, 92.2 on GAIA, and 93.8 on DeepSearchQA, and keeps improving as its budget of rounds grows. These results show that long-horizon reflective data is an effective route toward self-improving agents.

基于这些数据训练的智能体（基于 Qwen3.8-27B 构建）在 MLE-bench Lite (81.8) 和 Frontier-CS (70.7) 上取得了优异成绩，并成功迁移至深度研究任务，在 BrowseComp 上达到 84.0，HLE 上达到 52.6，GAIA 上达到 92.2，DeepSearchQA 上达到 93.8，且随着迭代轮数预算的增加，其性能持续提升。这些结果表明，长程反思数据是实现自我改进智能体的有效途径。

Paper Details:

Authors: Hongjin Qian, Chaofan Li, Kun Luo, Wenqing Wei, Jianlyu Chen, Shuqi Lu, Yuyang Hu, Hongwang Xiao, Hui Wang, Chaozhuo Li, Qiwei Ye, Zhicheng Dou, Defu Lian, Zheng Liu
arXiv ID: 2609.38288
Date: 29 Sep 2026

论文详情：

作者： Hongjin Qian, Chaofan Li, Kun Luo, Wenqing Wei, Jianlyu Chen, Shuqi Lu, Yuyang Hu, Hongwang Xiao, Hui Wang, Chaozhuo Li, Qiwei Ye, Zhicheng Dou, Defu Lian, Zheng Liu
arXiv ID: 2609.38288
日期： 2026年9月29日