[2603.01241] TARSE: Test-Time Adaptation via Retrieval of Skills and Experience for Reasoning Agents

Computer Science > Information Retrieval arXiv:2603.01241 (cs) [Submitted on 1 Mar 2026] Title:TARSE: Test-Time Adaptation via Retrieval of Skills and Experience for Reasoning Agents Authors:Junda Wang, Zonghai Tao, Hansi Zeng, Zhichao Yang, Hamed Zamani, Hong Yu View a PDF of the paper titled TARSE: Test-Time Adaptation via Retrieval of Skills and Experience for Reasoning Agents, by Junda Wang and 5 other authors View PDF HTML (experimental) Abstract:Complex clinical decision making often fails not because a model lacks facts, but because it cannot reliably select and apply the right procedural knowledge and the right prior example at the right reasoning step. We frame clinical question answering as an agent problem with two explicit, retrievable resources: skills, reusable clinical procedures such as guidelines, protocols, and pharmacologic mechanisms; and experience, verified reasoning trajectories from previously solved cases (e.g., chain-of-thought solutions and their step-level decompositions). At test time, the agent retrieves both relevant skills and experiences from curated libraries and performs lightweight test-time adaptation to align the language model's intermediate reasoning with clinically valid logic. Concretely, we build (i) a skills library from guideline-style documents organized as executable decision rules, (ii) an experience library of exemplar clinical reasoning chains indexed by step-level transitions, and (iii) a step-aware retriever that selects ...