Grounded treatment reasoning

ATHENA-R1

An AI agent for treatment reasoning over a biomedical tool universe.

Shanghua Gao1, Ayush Noori1,2,3, Richard Zhu1, Curtis Ginder1,4, Zhenglun Kong1, Xiaorui Su1, Justin Kauffman5, Benjamin S. Glicksberg5,6,7, Joshua Lampert5,6,8, Ankit Sakhuja5,9,10, Ashwin Sawant5,9,11, ATHENA-R1 Evaluation Consortium12, David A. Clifton2,13, Noa Dagan3,14,15, Ran Balicer3,14,16, Marinka Zitnik1,3,17,18,19,†
  1. Department of Biomedical Informatics, Harvard Medical School, Boston, MA
  2. Department of Engineering Science, University of Oxford, Oxford, UK
  3. The Ivan and Francesca Berkowitz Family Living Laboratory Collaboration at Harvard Medical School and Clalit Research Institute, Boston, MA, USA
  4. Cardiovascular Division, Department of Medicine, Brigham and Women's Hospital, Harvard Medical School, Boston, MA
  5. The Windreich Department of Artificial Intelligence and Human Health, Icahn School of Medicine at Mount Sinai, New York, NY, USA
  6. The Hasso Plattner Institute for Digital Health at Mount Sinai, Icahn School of Medicine at Mount Sinai and Mount Sinai Health System, New York City, NY, USA
  7. Mindich Child Health and Development Institute and the Departments of Pediatrics and Genetics & Genomic Sciences, Icahn School of Medicine at Mount Sinai, New York, NY, USA
  8. Mount Sinai Fuster Heart Hospital, Icahn School of Medicine at Mount Sinai, New York, NY, USA
  9. Mount Sinai AI Assurance Lab, Mount Sinai Health System, New York, NY, USA
  10. Institute for Critical Care Medicine, Icahn School of Medicine at Mount Sinai, New York, NY, USA
  11. Department of Medicine, Icahn School of Medicine at Mount Sinai, New York, NY, USA
  12. ATHENA-R1 Evaluation Group (the list of members and their affiliations appears in the Supplementary Information)
  13. Oxford Suzhou Centre for Advanced Research, University of Oxford, Suzhou, Jiangsu, China
  14. Clalit Research Institute, Innovation Division, Clalit Health Services, Ramat Gan, Israel
  15. Faculty of Computer and Information Science, Ben Gurion University of the Negev, Be'er Sheva, Israel
  16. Faculty of Health Sciences, School of Public Health, Ben Gurion University of the Negev, Be'er Sheva, Israel
  17. Kempner Institute for the Study of Natural and Artificial Intelligence, Harvard University, Cambridge, MA
  18. Broad Institute of MIT and Harvard, Cambridge, MA
  19. Harvard Data Science Initiative, Cambridge, MA

† Correspondence — marinka@hms.harvard.edu

reasoning trace · illustrative example, not clinical advice
▸ treatment problem
Which treatment adjustments are safe for metformin in an elderly patient with type 2 diabetes, hypertension, and early CKD?
65y maleeGFR 52ACE + HCTZlactic-acidosis concernonce-daily
— ready — step 0 / 23
TOOLS REASONING CLINICIAN parallel tools α | β | γ threads human input
▸ tool call◆ reasoning● clinician
◂ evidence-grounded answer
01Continue metformin — safe at eGFR 52 with monitoringopenFDA dosing label · HPO phenotype
02Reduce dose if eGFR drops below 45 mL/minFDA dosing · renal-clearance reasoning
03Switch ACE inhibitor → amlodipine 5 mg once dailyOpen Targets interaction · label safety
04Replace HCTZ → indapamide 1.25 mgdrug-label comparison
05Monitor eGFR + electrolytes every 3 monthsintegrated reasoning
0%

Reasoning, one grounded step at a time

Instead of answering in a single pass, ATHENA-R1 decides what evidence it needs, selects from a library of 212 biomedical tools, retrieves grounded evidence, and folds it into the next step — returning a final answer and a full reasoning trace. Built on a Qwen3-8B backbone.

Knowledge grounding

Retrieves verified evidence through tools and grounds every conclusion in source data — each step traces to a drug label or annotation.

Goal-oriented tool selection

Selects the most relevant tools per step from the 212-tool library, instead of packing every tool into context.

Multi-step reasoning

Decomposes a task, chains tool calls, interprets evidence and revises when results are incomplete or unexpected.

Real-time retrieval

Tools query continuously updated biomedical resources, reasoning over evidence beyond its parametric knowledge.

Treatment reasoning lies at the heart of medicine — weighing disease context, candidate therapies, comorbidities, contraindications and evolving evidence. It is inherently iterative: evidence must be gathered and revised across steps, not inferred in one pass.

We introduce ATHENA-R1, a grounded AI agent for treatment reasoning. At each step it identifies missing information, selects from 212 biomedical tools, retrieves evidence from curated sources, and folds it into subsequent reasoning. It is trained through two-level self-learning: multi-agent systems build tools, questions and reasoning traces for supervised fine-tuning, then reinforcement learning with rule-based scientific feedback.

Across five benchmarks (3,168 drug-reasoning tasks; 456 patient-specific scenarios) it leads language models and tool-use systems — 94.7% on open-ended drug reasoning (+17.8 pp over GPT-5, +25.9 pp over DeepSeek-R1) and 82.9% on TreatmentPC. Experts from 28 disease organizations prefer it across all eight criteria; in records from 5.4M patients, its adverse-event hypotheses track elevated risk in matched subpopulations.

A universe of 212 biomedical tools

Tools span drug mechanisms, interactions, safety and disease annotations, interfacing with openFDA (FDA labels since 1939), Open Targets and the Human Phenotype Ontology. The biomedical tool library is available as pip install tooluniverse. Check out ToolUniverse ↗ for an expanded library of general-purpose tools for scientific discovery beyond treatment reasoning.

Categories cycle automatically · click a wedge to pause
212 biomedical tools
Tool category

    Two-level self-learning

    Reasoning traces are too large and varied to annotate by hand. ATHENA-R1 learns them from generated traces — first the structure of reasoning, then how to act within it.

    DataGen multi-agent pipeline · Level 1
    ToolGen212 tools from APIs QuestionGen85,340 tasks TRACEGEN Helper Tool Provider Solver ATHENA-R1-Instruct378,027 SFT samples ↻ iterative
    The DataGen multi-agent system builds the tools, writes tasks from FDA labels, and composes step-by-step reasoning traces.
    378,027
    Instruction samples
    85,340
    Reasoning traces
    177,626
    Reasoning steps
    281,695
    Tool calls
    RL with scientific feedback · Level 2
    PolicyATHENA-R1 Rollouts in212-tool env SCIENTIFIC FEEDBACK · 6 1 · correctness 2 · format 3 · evidence 4 · multi-step 5 · arg grounding 6 · non-redundant GRPO update
    Reinforcement learning: rollouts are scored on six rule-based dimensions, and GRPO favors higher-scoring traces.
    TreatmentPC · ablation
    39.2%
    Qwen3-8B base
    +27.3
    66.5%
    + Level 1 · SFT
    +8.3
    74.8%
    + Level 2 · RL
    Both levels matter: supervised fine-tuning lifts the base by 27.3 pp and reinforcement learning adds a further 8.3 pp.

    Beating frontier reasoning models

    Across five datasets, ATHENA-R1 leads in open-ended evaluation. DrugPC uses FDA drugs approved in 2024 — held out of training — to limit leakage.

    DrugPC · open-ended accuracy
    ATHENA-R194.7
    GPT-576.9
    DeepSeek-R1 671B68.8
    Qwen348.7
    3,168 questions · 11 categories — +17.8 pp over GPT-5
    Open-ended accuracy on DrugPC across 11 categories of drug information.
    TreatmentPC · open-ended accuracy
    ATHENA-R182.9
    GPT-572.2
    DeepSeek-R1 671B67.5
    Qwen3-Next60.1
    Qwen339.2
    456 patient-specific scenarios — +10.7 pp over GPT-5
    On TreatmentPC the answer depends on patient context. Giving GPT-5 tool access doesn't close the gap — ATHENA-R1 uses tools on every problem.

    Preferred by rare-disease experts

    With the Chan Zuckerberg Initiative Rare As One network — 29 experts from 28 disease organizations — ATHENA-R1 was compared blind against Qwen3, GPT o3-mini, Gemini-2.0-Flash and DeepSeek-R1. 23 evaluators gave 110 responses; experts preferred it across all eight criteria.

    Blinded preference · 8 criteria
    reference model preferredpairwise preference · % of 110 responsesATHENA-R1 preferred
    3.6
    Cognitive traceability
    95.5
    2.7
    Helpfulness of rationale
    94.5
    17.3
    Completeness
    66.4
    19.1
    Task success
    63.6
    19.1
    Possibility of harm
    61.8
    16.4
    Alignment w/ consensus
    59.1
    18.2
    Accuracy of content
    58.2
    20.0
    Clinical relevance
    57.3
    Ties and "neither did well" responses (10–25% combined) not shown on bars.
    4.16
    ATHENA-R1 · mean ±0.90
    2.44
    reference models ±1.26
    Share of evaluations preferring ATHENA-R1, and mean absolute ratings (1–5). All differences significant (P < 5×10⁻⁵). Physicians separately rated five real clinical vignettes — task success averaged 4.63 ± 0.52.

    Risk hypotheses tested in 5.4M patients

    These hypotheses are AI-generated, not literature-derived — ATHENA-R1 reasoned over disease + comorbidity + medication profiles to predict adverse events that should be elevated in specific patient subpopulations. The predictions were then tested retrospectively against records from over 5.4 million patients at Clalit Health Services, using matched cohort analyses adjusted for age, sex, socioeconomic status and outpatient utilization.

    Adjusted odds ratios · 95% CI
    1.00.51.52.0 ADJUSTED ODDS RATIO1 = no effect · >1 = elevated risk in targeted subpopulation β-blocker → acute kidney failurehypertension + gout 1.84 β-blocker → hyperkalemiahypertension + gout 1.78 DPP-4 inh. → hepatocellular carcinomadiabetes + ischemic heart disease 1.48 diuretic → squamous cell carcinomahypertension + gout 1.08 statin → liver failurehyperlipidemia + hypothyroidism 1.04 metformin → respiratory failurediabetes + chronic kidney disease 1.00
    Three of six AI-generated predictions reached statistical significance after adjusting for confounders, with odds ratios of 1.48–1.84 in the targeted subpopulations. Red = confirmed (95% CI excludes 1); gray = adjusted regression did not confirm in this cohort, though prevalence patterns aligned for two of the three.

    Run it. Cite it.

    # install the tool library pip install tooluniverse # code · demos · weights github.com/mims-harvard/ATHENA huggingface.co/mims-harvard/ATHENA-R1-8B
    @article{gao2026athena,
      title  = {An AI agent for treatment reasoning
                 over a biomedical tool universe},
      author = {Gao, Shanghua and Noori, Ayush and
                 Zhu, Richard and Ginder, Curtis and
                 Kong, Zhenglun and Su, Xiaorui and
                 Kauffman, Justin and Glicksberg, Benjamin S. and
                 Lampert, Joshua and Sakhuja, Ankit and
                 Sawant, Ashwin and
                 {ATHENA-R1 Evaluation Consortium} and
                 Clifton, David A. and Dagan, Noa and
                 Balicer, Ran and Zitnik, Marinka},
      year   = {2026}
    }

    Questions? Email Shanghua Gao and Marinka Zitnik.