{"ID":5675996,"CreatedAt":"2026-07-03T01:40:09.565152011Z","UpdatedAt":"2026-07-04T20:51:25.697068714Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2607.01470","arxiv_id":"2607.01470","title":"World Feedback for Clinical Agents: Diagnosing RL in FHIR Environments","abstract":"Clinical protocol-execution tasks -- checking a lab value, applying a threshold, placing a correctly structured FHIR order -- are natural candidates for RL from world feedback: once clinical SMEs encode decision logic into a verifier, that verifier grades unlimited rollouts without per-episode annotation. But applying RL requires a sound feedback channel and sufficient base capability. We audit MedAgentBench v1/v2, find a 41.7\\% silent-finish ceiling that makes inaction the RL dominant strategy, and construct \\textbf{MedAgentBench-v3 (MAB-v3)} (508 tasks, 8.9\\% ceiling). Training Qwen3-8B exposes two structural barriers: a \\emph{capability ceiling} (10/20 task types have 0\\% base performance, zero gradient) and a \\emph{format-knowledge barrier} (3/20 types require exact clinical codes undiscoverable by exploration). Pure RL reaches 18.2\\% pass@1 vs.\\ 34.1\\% for rule-based SFT; the 15.9~pp gap is attributable entirely to these barriers. A decision/format-knowledge/lookup taxonomy predicts RL learnability and prescribes the fix: SFT to inject codes, RL to learn conditionals.","short_abstract":"Clinical protocol-execution tasks -- checking a lab value, applying a threshold, placing a correctly structured FHIR order -- are natural candidates for RL from world feedback: once clinical SMEs encode decision logic into a verifier, that verifier grades unlimited rollouts without per-episode annotation. But applying...","url_abs":"https://arxiv.org/abs/2607.01470","url_pdf":"https://arxiv.org/pdf/2607.01470v1","authors":"[\"Ananya Mantravadi\",\"Harshit Rajgarhia\",\"Prasanna Desikan\",\"Abhishek Mukherji\"]","published":"2026-07-01T21:02:54Z","proceeding":"cs.AI","tasks":"[\"cs.AI\"]","methods":"[\"LoRA\"]","has_code":false}