Traj-Evolve: A Self-Evolving Multi-Agent System for Patient Trajectory Modeling in Lung Cancer Early Detection
Traj-Evolve is a self-evolving multi-agent system designed for modeling patient trajectories in lung cancer early detection. It utilizes an Experience Pool and multi-agent reinforcement learning to enhance prediction accuracy. The system outperforms existing models by effectively leveraging historical patient data.
- ▪Traj-Evolve addresses the challenges of modeling patient trajectories from electronic health records.
- ▪The system incorporates an Experience Pool for retrieving similar patient cases as few-shot contexts.
- ▪It outperforms nine strong baselines in lung cancer prediction tasks, particularly in a challenging never-smoker population.
Opening excerpt (first ~120 words) tap to expand
Computer Science > Artificial Intelligence arXiv:2606.02812 (cs) [Submitted on 1 Jun 2026] Title:Traj-Evolve: A Self-Evolving Multi-Agent System for Patient Trajectory Modeling in Lung Cancer Early Detection Authors:Sihang Zeng, Matthew Thompson, Ruth Etzioni, Meliha Yetisgen View a PDF of the paper titled Traj-Evolve: A Self-Evolving Multi-Agent System for Patient Trajectory Modeling in Lung Cancer Early Detection, by Sihang Zeng and 3 other authors View PDF HTML (experimental) Abstract:Modeling patient trajectories from longitudinal electronic health records (EHRs) requires reasoning over sparse, noisy, and long-context multimodal sequences.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at arXiv cs.AI.
Discussion
0 commentsMore from arXiv cs.AI
