الملخص
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.
الكلمات المفتاحية
الموضوع
بيانات النشر
- المجلة
- غير متاح
- وصول مفتوح
- وصول مفتوح أخضر
اقتبس هذه المقالة
APA 7
Qian, H., Li, C., Luo, K., Wei, W., Chen, J., Lu, S., Hu, Y., Xiao, H., Wang, H., Li, C., Ye, Q., Dou, Z., Lian, D., & Liu, Z. (2026). AREX-2: Advancing Self-Improving Agents through Long-Horizon Reflective Tasks. https://omanscience.com/ar/articles/arex-2-advancing-self-improving-agents-through-long-horizon-reflective-tasks
MLA 9
Qian, Hongjin, et al. "AREX-2: Advancing Self-Improving Agents through Long-Horizon Reflective Tasks." https://omanscience.com/ar/articles/arex-2-advancing-self-improving-agents-through-long-horizon-reflective-tasks.
شيكاغو (المؤلف–التاريخ)
Qian, Hongjin, Chaofan Li, Kun Luo, Wenqing Wei, Jianlyu Chen, Shuqi Lu, Yuyang Hu, Hongwang Xiao, Hui Wang, Chaozhuo Li, Qiwei Ye, Zhicheng Dou, Defu Lian, and Zheng Liu. 2026. "AREX-2: Advancing Self-Improving Agents through Long-Horizon Reflective Tasks." https://omanscience.com/ar/articles/arex-2-advancing-self-improving-agents-through-long-horizon-reflective-tasks.
هارفارد
Qian, H., Li, C., Luo, K., Wei, W., Chen, J., Lu, S., Hu, Y., Xiao, H., Wang, H., Li, C., Ye, Q., Dou, Z., Lian, D. and Liu, Z. (2026) 'AREX-2: Advancing Self-Improving Agents through Long-Horizon Reflective Tasks', Available at: https://omanscience.com/ar/articles/arex-2-advancing-self-improving-agents-through-long-horizon-reflective-tasks.
فانكوفر
Qian H, Li C, Luo K, Wei W, Chen J, Lu S, et al. AREX-2: Advancing Self-Improving Agents through Long-Horizon Reflective Tasks. https://omanscience.com/ar/articles/arex-2-advancing-self-improving-agents-through-long-horizon-reflective-tasks
IEEE
H. Qian, C. Li, K. Luo, W. Wei, J. Chen, S. Lu, Y. Hu, H. Xiao, H. Wang, C. Li, Q. Ye, Z. Dou, D. Lian, and Z. Liu, "AREX-2: Advancing Self-Improving Agents through Long-Horizon Reflective Tasks," https://omanscience.com/ar/articles/arex-2-advancing-self-improving-agents-through-long-horizon-reflective-tasks.