analiza aiqbench
This research addresses the critical challenge of generalization in autonomous agents by merging supervised fine-tuning with episodic retrieval mechanisms. The study demonstrates that integrating retrieved trajectories directly into the fine-tuning process significantly enhances performance on unseen tasks, even when leveraging imperfect or self-generated experience. This framework offers a practical pathway for building more adaptable AI agents without relying solely on extensive parameter updates at test time.
Retrieval-Augmented LLM Agents: Learning to Learn from Experience
arXiv:2603.18272v2 Announce Type: replace Abstract: While large language models (LLMs) have advanced the development of general-purpose agents, robust generalization to unseen tasks remains challenging. Two common approaches are supervised fine-tuning and training-free memory-augmented generation using retrieved experience; yet both have limitations: fine-tuning often fails to extrapolate to new tasks, while experience retrieval often underperforms compared to supervised baselines. In this work, we combine these approaches and study how retrieval-augmented LLM agents can learn to use retrieved trajectories in-context. First, we establish a strong LoRA fine-tuning baseline that outperforms several state-of-the-art agent training pipelines. Second, we analyze key design choices for experience retrieval, including storage, querying, and trajectory selection. We then integrate experience retrieval directly into the fine-tuning process, finding that this substantially improves generalization to unseen tasks. Finally, we show that these gains often persist with imperfect experience and, even when agents reuse their own failed attempts without test-time parameter updates. Overall, our results establish simple episodic retrieval as a strong foundation for agent memory and retrieval-aware fine-tuning as a practical and effective framework for building agents that learn to learn from experience.
Abstract:While large language models (LLMs) have advanced the development of general-purpose agents, robust generalization to unseen tasks remains challenging. Two common approaches are supervised fine-tuning and training-free memory-augmented generation using retrieved experience; yet both have limitations: fine-tuning often fails to extrapolate to new tasks, while experience retrieval often underperforms compared to supervised baselines. In this work, we combine these approaches and study how retrieval-augmented LLM agents can learn to use retrieved trajectories in-context. First, we establish a strong LoRA fine-tuning baseline that outperforms several state-of-the-art agent training pipelines. Second, we analyze key design choices for experience retrieval, including storage, querying, and trajectory selection. We then integrate experience retrieval directly into the fine-tuning process, finding that this substantially improves generalization to unseen tasks. Finally, we show that these gains often persist with imperfect experience and, even when agents reuse their own failed attempts without test-time parameter updates. Overall, our results establish simple episodic retrieval as a strong foundation for agent memory and retrieval-aware fine-tuning as a practical and effective framework for building agents that learn to learn from experience.
Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
arXiv:2603.18272 [cs.AI]
(or arXiv:2603.18272v2 [cs.AI] for this version)
https://doi.org/10.48550/arXiv.2603.18272 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 274 cuvinte
Subiecte conexe
CLEAR: Context Augmentation from Contrastive Learning of Experience via Agentic Reflection
arXiv · 16 sept. 2026
What's Continual Learning, and Why Might We Expect To See It In Advanced LLM Agents?
LessWrong · 12 iun. 2026
VikingRAG: Accurate and Token-efficient Retrieval-augmented Generation over Structured Documents
arXiv · 12 sept. 2026
GenMem: Generative Symbolic Memory for Self-Evolving Harness
arXiv · 30 sept. 2026