Search papers, labs, and topics across Lattice.
This paper introduces Learning to Retrieve from Agent Trajectories (LRAT), a novel training paradigm for retrieval models used by LLM-powered agents. LRAT leverages multi-step agent interactions to derive supervision signals, including browsing actions, unbrowsed rejections, and post-browse reasoning traces, addressing the mismatch between human-centric retrieval models and agent query patterns. Experiments on research benchmarks demonstrate that LRAT-trained retrievers improve evidence recall, task success, and efficiency across various agent architectures and scales.
Forget human clicks: training retrieval models directly from agent behavior unlocks significant gains in task success and efficiency for LLM-powered search agents.
Information retrieval (IR) systems have traditionally been designed and trained for human users, with learning-to-rank methods relying heavily on large-scale human interaction logs such as clicks and dwell time. With the rapid emergence of large language model (LLM) powered search agents, however, retrieval is increasingly consumed by agents rather than human beings, and is embedded as a core component within multi-turn reasoning and action loops. In this setting, retrieval models trained under human-centric assumptions exhibit a fundamental mismatch with the way agents issue queries and consume results. In this work, we argue that retrieval models for agentic search should be trained directly from agent interaction data. We introduce learning to retrieve from agent trajectories as a new training paradigm, where supervision is derived from multi-step agent interactions. Through a systematic analysis of search agent trajectories, we identify key behavioral signals that reveal document utility, including browsing actions, unbrowsed rejections, and post-browse reasoning traces. Guided by these insights, we propose LRAT, a simple yet effective framework that mines high-quality retrieval supervision from agent trajectories and incorporates relevance intensity through weighted optimization. Extensive experiments on both in-domain and out-of-domain deep research benchmarks demonstrate that retrievers trained with LRAT consistently improve evidence recall, end-to-end task success, and execution efficiency across diverse agent architectures and scales. Our results highlight agent trajectories as a practical and scalable supervision source, pointing to a promising direction for retrieval in the era of agentic search.
