Toward TheraAgent: Evidence-Grounded, Self-Verifying AI for Oncology Treatment Recommendation

Junhan Wang, David Scott Lewis, Enrique Zueco

Submission Type: I want my submission to be considered for both oral and poster presentation.

Keywords: oncology treatment recommendation, AI for medicine and healthcare, agents and autonomous AI researchers, clinical decision support, evidence-grounded AI, self-verifying AI, agentic workflows, multi-agent systems, retrieval-augmented generation, human-in-the-loop AI, trustworthy AI, uncertainty-aware AI, abstention-aware AI, guideline-grounded reasoning, patient-state abstraction, multimodal evidence retrieval, treatment sequencing, radiation oncology, biomarker-informed care, provenance, healthcare governance, evaluation benchmarks, translational oncology, auditable AI

TL;DR: TheraAgent proposes an oncology copilot that abstracts patient state, retrieves evidence in steps, drafts ranked options, and independently verifies each claim so unsupported recommendations become abstentions before clinician review.

Confirmation Of Submission Requirements: I submit an abstract. It uses the template provided on the submission page and is no longer than 2 pages.

Submission Number: 405