An AI agent for treatment reasoning over a biomedical tool universe

2026-06-26

AI Summary

This work addresses the challenge of modeling evidence-driven, dynamic therapeutic reasoning that integrates multidimensional clinical information—a task poorly suited to conventional AI approaches. The authors propose ATHENA-R1, an agent-based system featuring a two-stage self-learning framework: it first employs multi-agent collaboration to generate training trajectories for supervised fine-tuning, then refines its reasoning through reinforcement learning guided by scientific feedback, achieving the first fully unsupervised training paradigm for treatment reasoning. Operating within an environment equipped with 212 biomedical tools, ATHENA-R1 autonomously identifies information gaps, invokes relevant tools, and synthesizes evidence. Evaluated on 3,168 drug-reasoning tasks and 456 real-world cases, it achieves 94.7% accuracy in open-ended reasoning and 82.9% in therapeutic decision-making—significantly outperforming GPT-5. Expert blind assessments consistently favor ATHENA-R1, and its adverse drug reaction hypotheses are validated across 5.4 million electronic health records.

Abstract

Treatment reasoning underpins every therapeutic decision, integrating disease context, comorbidities, medications, contraindications, and evolving biomedical knowledge to select an appropriate therapy. It is inherently iterative: candidates are weighed against many constraints, revised as evidence emerges, and grounded in verifiable sources. Here we introduce ATHENA-R1, an AI agent for treatment reasoning across all FDA approved drugs since 1939, trained by reinforcement learning over a universe of 212 biomedical tools. At each step it identifies missing information, selects and runs relevant tools, and incorporates the evidence. To train it without human-annotated traces, we build a two-level self-learning framework: multi-agent systems construct the tools, tasks, and reasoning trajectories for supervised fine-tuning, then reinforcement learning with scientific feedback rewards reasoning quality (evidence gathering, grounded tool use, logical non-redundancy). Across five benchmarks of 3,168 drug reasoning tasks and 456 patient treatment cases, ATHENA-R1 outperforms language models and tool-use systems, reaching 94.7% accuracy on open-ended drug reasoning and 82.9% on treatment reasoning, 17.8 and 10.7 points above GPT-5. In blinded evaluations by experts from 28 rare disease organizations, it is preferred over reference models on all criteria, and physicians rated it favorably on complex hospitalized cardiovascular and infectious-disease cases. Adverse-event hypotheses it generated, tested in electronic health records from 5.4 million patients, reached adjusted odds ratios of 1.48-1.84, with no elevation among negative controls. Because it requires knowing what evidence to seek before concluding, treatment reasoning has long been hard for AI; we show it can be reframed as a learnable process of iterative evidence gathering that reinforcement learning can train AI to perform.

Problem

• treatment reasoning
• biomedical tools
• evidence gathering
• AI agent
• drug therapy

Innovation

• treatment reasoning
• reinforcement learning
• biomedical tool use
• self-learning framework
• iterative evidence gathering