04:00
2026-10-01
arxiv.org
ai-agents
AREX-2: Advancing Self-Improving Agents through Long-Horizon Reflective Tasks
AREX-2, an LLM agent built on Qwen3.8-27B and trained on synthesized long-horizon improvement trajectories from machine learning and algorithmic programming tasks, scored 81.8 on MLE-bench Lite and 70…