Sage-Taiyi leads Taiyi-Bench scientific ideation scores
Original title: Sage-Taiyi: A Metacognitive Architecture for Scientific Discovery
It scores 54.0 overall versus 47.9 for AI Scientist-v2, and authors report chemistry yield gains in collaborations.
Researchers introduce Sage-Taiyi, a metacognitive system that revises evidence, assumptions, and open questions to choose research targets and reusable methods. On Taiyi-Bench, 265 tasks from 50 papers across ten disciplines, they say it ranks first on all eight historical-rediscovery and open-ended ideation metrics, with an Overall score of 54.0 against 47.9 for AI Scientist-v2. Five expert collaborations link proposals to validation; reported chemistry yields rise from trace to 84% and from 13% to 69%.
Unconfirmed: Benchmark ranks and yield figures are author-reported and not independently confirmed in this abstract.
- Taiyi-Bench has 265 tasks drawn from 50 papers in ten disciplines.
- Sage-Taiyi also does well on two external ideation benchmarks, the authors say.
- Cobalt-catalyzed carbonylation yield is reported rising from trace amounts to 84%.
- Photocatalytic allylic amination yield is reported rising from 13% to 69%.
- Products
- Taiyi-Bench
- AI models
- Sage-TaiyiAI Scientist-v2
From the original article
1 min read at arxiv.orgWe introduce Sage-Taiyi, a metacognition-inspired architecture that uses accumulated scientific understanding to guide what to investigate next and how to investigate it. Through explicit monitoring and control, the system revises its understanding of evidence, assumptions, and unresolved questions, then uses these assessments to adapt research targets and select reusable reasoning methods.
By Weida Wang · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.