Sage-Taiyi: A Metacognitive Architecture for Scientific Discovery
Sage-Taiyi scored 54.0 against the strongest baseline’s 47.9 on the authors’ scientific ideation benchmark.
The system monitors evidence, assumptions and unresolved questions to adjust research targets and choose reusable reasoning methods. On Taiyi-Bench’s 265 tasks across ten disciplines, the authors report that it ranked first on all eight metrics. Five expert-guided collaborations connected its proposals to experimental and computational validation. Reported chemistry results included raising cobalt-catalyzed carbonylation yield from trace amounts to 84% and photocatalytic allylic amination yield from 13% to 69%.
ArXiv · AI/CL/LG's note
The system monitors evidence, assumptions and unresolved questions to adjust research targets and choose reusable reasoning methods. On Taiyi-Bench’s 265 tasks across ten disciplines, the authors report that it ranked first on all eight metrics. Five expert-guided collaborations connected its proposals to experimental and computational validation. Reported chemistry results included raising cobalt-catalyzed carbonylation yield from trace amounts to 84% and photocatalytic allylic amination yield from 13% to 69%.
ArXiv · AI/CL/LG's note
score 5