发表机构
Beijing Academy of Artificial Intelligence (BAAI)（北京人工智能研究院）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
AREX-2通过合成长时程反思轨迹训练LLM智能体，提升其测试时迭代优化能力，在多个基准上取得显著成绩并持续改进。
AI 中文摘要
我们提出AREX-2，旨在推进LLM智能体的自我改进能力，我们将其定义为在测试时迭代优化解决方案的能力。这种能力依赖于两种互补的能力：反思，即产生比当前方案更好的解决方案；以及长时程执行，即保持迭代在多轮中有效。我们假设这两种能力都是领域无关的，因此可以在适合监督的场景中学习。据此，我们从机器学习和算法编程任务中合成长时程改进轨迹，这两个领域提供可验证的反馈并奖励持续迭代。基于这些数据训练，我们的智能体基于Qwen3.8-27B构建，在MLE-bench Lite（81.8）和Frontier-CS（70.7）上取得了强劲结果，并迁移到深度研究领域，在BrowseComp上达到84.0，在HLE上达到52.6，在GAIA上达到92.2，在DeepSearchQA上达到93.8，并且随着轮次预算的增长而持续改进。这些结果表明，长时程反思数据是通往自我改进型智能体的有效途径。
英文摘要
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. 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. 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.
CommentsCode will be released at https://github.com/VectorSpaceLab/AREX-2 and models at https://huggingface.co/collections/BAAI/arex-2
