The paper introduces Cura 1T, a healthcare-specialized language model designed to cover patient consultation, clinical reasoning over text and images, interactive diagnosis, and electronic health record tool use. Its central method is a human-gated self-evolution loop: a training agent selects a capability target, trains the model, evaluates benchmark trajectories, and revises the data mixture based on observed failures. The abstract reports that Cura 1T ranks at or near the top among frontier baselines on a healthcare evaluation suite while remaining competitive on out-of-domain reasoning and agentic benchmarks. It does not provide benchmark names or numerical results in the supplied summary.
No heat snapshots are available in the last 24 hours.