Sage-Taiyi: A Metacognitive Architecture for Scientific Discovery
AuthorsWeida Wang, Benteng Chen, Yigeng Jiang, Shuchen Pu, Weiping Shen, Zichao Liu, Yulin Su, Zitong Wang, Siyu Zhou, Zifu Wang, Yizhi Li, Xin Zong, Yuning Yang, Jixiang Bi, Yuan Wang, Chuyi Peng, Shiyao Geng, Qi Zhang, Kuan Lu, Mengzhu Li, Weiwang Xu, Peishan Xiang, Chengjun Liu, Chengrui Wei, Fangchen Yu, Shengji Tang, Yichen Jiang, Jingqi Ye, Cheng Yang, Tao Hu, Jianbiao Mei, Qinglong Cao, Kun Li, Yixin Chen, Shengyuan Xu, Juncheng Xiao, Yiheng Wang, Ruizhe Chen, Shiyang Feng, Runmin Ma, Xiangchao Yan, Zongsheng Cao, Haomeng Hu, Yu Ke, Taihui Wu, Meng Dai, Xu Guo, Haoyang Peng, Yinglong Ni, Pingsen Shi, Xiaosong Wang, Shuyue Hu, Haijun Li, Sherry Dongqi Bao, Yang Yang, Chao Yang, Jie Hou, Tianfan Fu, Yi Gong, Jin Zeng, Mao Su, Hongxia Hao, Fanyang Mo, Fenghua Ling, Tianshu Yu, Yunqi Cai, Baihong Chi, Jingwei Wang, Lei Wang, Chuncheng Bu, Shide Lv, Guangyu Wang, Yuqiang Li, Chunfeng Song, Lei Bai, Hui Wang, Xujun Fu, Li Cao, Lei Chen, Dahua Lin, Xing Xie, Bowen Zhou, Wenlong Zhang, Pinl...
Sage-Taiyi uses metacognitive reasoning to choose scientific questions and methods, with benchmark gains and promising results in laboratory collaborations.
Key results
Benchmark tasks drawn from 50 papers across ten disciplines.
Taiyi-Bench score; highest among evaluated systems.
Strongest baseline on Taiyi-Bench.
Optimized yield, increased from trace amounts.
Yield before reaction-condition optimization.
Yield after reaction-condition optimization.
What the paper found
Sage-Taiyi is a scientific-ideation system that explicitly monitors its evolving understanding and uses it to decide what to investigate next and which reasoning methods to apply. Its object-level search organizes research targets, idea lineages, and evidence; a meta-level Cognitive State tracks support, uncertainty, assumptions, and shared dependencies, while a reusable repertoire supplies research strategies. On Taiyi-Bench, covering 265 tasks from 50 papers across ten disciplines, Sage-Taiyi scores 54.0 overall, ahead of AI Scientist-v2 at 47.9 and first on all eight metrics. The evaluation compares systems built around models and tools including OpenAI’s GPT-5.6-Sol and Codex, Anthropic’s Claude Opus 5 and Claude Code, and Google’s Gemini 3.7 Flash. Five expert-guided collaborations connect proposals to experiments or simulations: reaction-condition optimization raises cobalt-catalyzed carbonylation from trace yield to 84%, and photocatalytic allylic amination from 13% to 69%. The central contribution is a control loop that carries evidence-grounded lessons between search branches, so findings can redirect targets, revise assumptions, and guide practical validation.
Original abstract
We 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. On Taiyi-Bench, an AI co-scientist ideation benchmark comprising 265 tasks from 50 papers across ten disciplines, Sage-Taiyi ranks first on all eight metrics of historical rediscovery and open-ended ideation, achieving an Overall score of 54.0 compared with 47.9 for the strongest baseline, AI Scientist-v2. It also performs well on two external scientific ideation benchmarks. Five expert-guided collaborations further connect its proposals to experimental and computational validation across chemistry, materials science, and engineering. In chemistry, reaction-condition optimization raises the yield of cobalt-catalyzed carbonylation from trace amounts to 84% and photocatalytic allylic amination from 13% to 69%. These results support metacognitive regulation as an approach to scientific ideation and its progression toward practical validation.
Read the original paperMore in AI for Science
Browse all 58 papers →DrugCoding: A Scientific Programming Agent for Drug Discovery
Junwei Hu, Zhuohao Lin, Kun Li, Hengnan Zhang
DrugCoding is an AI agent that plans and executes multi-step computational tasks to help automate drug discovery.
Reliability by Architecture: Agentic Workflows for Validated Multi-Physics Simulation from Natural Language
Søren Taverniers, Eric Darve, Hamdi A. Tchelepi
An LLM-assisted workflow prepares and repairs multi-physics simulations, while deterministic checks help ensure the results match both the physics and the user’s request.
Sample-Efficient Control of Magnetoconvective Flows via Differentiable Simulation
Jannis Becktepe, Dmitry Krasnov, Nils Thuerey, Jörg Schumacher, Sebastian Peitz
A differentiable simulator helps researchers control magnetoconvective flows more efficiently and design better heat-transfer outcomes.