Search papers, labs, and topics across Lattice.
This paper introduces CASE, a framework for training clinical agents to autonomously gather evidence from longitudinal medical records and imaging, addressing limitations of existing benchmarks that rely on preselected data. By employing a novel agentic post-training approach with supervised fine-tuning and reinforcement learning, the system learns to strategically select relevant visits, modalities, and tools for investigation. The proposed approach significantly enhances evidence-grounded longitudinal reasoning, outperforming leading LLMs on a new UK Biobank-derived multimodal benchmark.
Clinical LLM agents can now autonomously gather evidence from longitudinal medical records and imaging, achieving superior accuracy to GPT-5.4 and Claude Opus.
Foundation models can serve as clinical agents through tool-use harnesses. However, conventional medical benchmarks assess reasoning over preselected evidence rather than the ability to seek it across clinical records and longitudinal imaging. We propose CASE: a series of role-specific Clinical Agents for Seeking Evidence, together with a tool-use harness and an agentic post-training framework for compact vision-language policy models. We further introduce a longitudinal multimodal benchmark built on UK Biobank, comprising 50,401 clinical questions derived from real-world ICD-10-coded diagnoses of 4,739 participants. Each question links to a patient-specific environment containing clinical context and multi-sequence MRI from baseline and follow-up visits, where agents autonomously select which visits, organs, modalities, slices, and specialist tools to inspect and compare. Supervised fine-tuning transfers evidence-seeking workflows from 14,734 frontier-model interaction trajectories, followed by agentic reinforcement learning on the learner's own environment interactions. Privileged on-policy self-distillation and rubric-based LLM feedback refine evidence-to-conclusion reasoning without prescribing tool sequences. Experiments show that CASE moves beyond question-answer imitation toward transferable investigation policies, strengthening evidence-grounded longitudinal reasoning. Under matched evaluation conditions, our Qwen3-VL-8B based agent achieves over 16% and 10% relative improvements in answer accuracy over GPT-5.4 and Claude Opus 4.8. Code will be available at https://github.com/VinyehShaw/CASE.
DAMO