Search papers, labs, and topics across Lattice.
This paper introduces an agentic AI Scientist workflow that integrates literature-guided reasoning, automated code generation, and hypothesis-driven experimentation to streamline the development of competitive deep learning baselines in medical imaging. By evaluating this framework across four public benchmarks, the authors achieve notable leaderboard placements, including 6th on the PUMA tracks and 31st on MILK10k, while also demonstrating strong domain generalization on MIDOG25. The results indicate that this skill-based approach significantly reduces the engineering effort typically required for baseline development in medical imaging tasks.
Automating the entire medical imaging model-development pipeline can yield competitive baselines with minimal engineering effort, challenging traditional iterative approaches.
Developing competitive deep learning baselines for medical imaging remains a highly iterative process requiring literature review, implementation, experimentation, and expert refinement. Existing automation approaches typically optimize isolated components, such as architecture search or hyperparameter tuning, rather than the complete baseline development process. We present an agentic AI Scientist workflow that combines literature-guided reasoning, automated code generation, and hypothesis-driven experimentation to generate competitive baseline models for medical imaging challenges. The framework is evaluated on four public benchmarks spanning segmentation, classification, and detection. Across all tasks, the Experimentation Pipeline consistently improves validation performance, achieving competitive leaderboard results, including 6th place on both PUMA tracks (15 teams) and 31st place on MILK10k (125 teams). On MIDOG25, the resulting model also demonstrates strong domain generalization across scanners, tumor types, and species. Using the same workflow across all challenges without task-specific redesign, we demonstrate that skill-based, literature-guided agentic workflows can substantially reduce the engineering effort required to develop competitive medical imaging baselines.
