Abstract:Recent multi-LLM agent systems have shown promising capabilities for automated problem-solving, yet they predominantly rely on frozen agents or static fine-tuning pipelines. To address this limitation, our primary contribution is ATLAS (Adaptive Task-distributed Learning for Agentic Self-evolution), a multi-agent framework where specialized meta-agents collaboratively train and refine an active agent toward a domain-specific policy. A core challenge in iterative preference learning within these pipelines is the reliance on fixed reference models, which typically leads to overly conservative updates or training stagnation. To overcome this, the framework's algorithmic engine utilizes Evolving Direct Preference Optimization (EvoDPO). EvoDPO employs an inspection agent to perform adaptive, proxy-KL gated reference policy updates based on continuous training telemetry. We evaluate this full framework across a diverse set of challenging environments-including non-stationary contextual bandits, partial differential equations (PINNs), and combinatorial optimization tasks (TSP, Bin Packing). Through comparison against fixed-reference, adaptive-reference, and external automated-discovery baselines, our results suggest that ATLAS combines supporter-driven exploration with EvoDPO-driven stability to improve long-horizon evaluator-driven self-improvement.
| Subjects: | Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2602.02709 [cs.AI] |
| (or arXiv:2602.02709v3 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2602.02709 arXiv-issued DOI via DataCite |
From: Ujin Jeon [view email]
[v1]
Mon, 2 Feb 2026 19:23:33 UTC (456 KB)
[v2]
Thu, 12 Feb 2026 22:37:34 UTC (456 KB)
[v3]
Thu, 21 May 2026 13:59:01 UTC (175 KB)
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。