OmniScientist: An Omni-Modal Omni-Discipline AI Scientist

Summary

Existing AI scientist systems typically reason over text, code, or precomputed summaries, missing crucial spatial, temporal, and cross-channel evidence. To address this, researchers introduce OmniScientist, an end-to-end omni-modal AI scientist that directly conducts multidisciplinary research from heterogeneous raw evidence. The system features a perception layer and three autonomous agents for ideation, experiment, and writeup, operating within a deterministic pipeline, allowing observations to shape research throughout the lifecycle. It enforces novelty, rigor, and claim checks via code. Evaluated on 36 real-data cases spanning 5 discipline families and 4 evidence types, including images, signals, and 3-D structures, it completed full pipelines in all cases, achieving a mean paper score of 6.3. Compared to a blind variant using precomputed scalar features, direct perception improved all 7 evaluation dimensions and won 85% of head-to-head judgments.