arXiv:2507.20230 · v4
Building verifiable databases at scale from the chemical literature via a multi-agent system
摘要
Accelerating chemical research with artificial intelligence requires high-quality databases. However, many reactions exist only as graphics drawn for human readers, which machines must read exactly. Specialist software is precise but rigid, whereas multimodal language models are flexible yet often write incorrect structures; neither flags errors. We present ChemEAGLE, a multi-agent system where tools supply structures and a language model extends their reach through judgments that programs apply. On 424 published graphics, its F1 score for exactly correct reactions was 82.2%, against 47.7% for a tool-equipped generalist agent, and calibrated confidence limited review to 35% of reactions at 96.6% accuracy. Without review, it built a 60,103-reaction photochemistry database whose records trained models to rank conditions and predict products. Verifiable databases can thus be built at scale.
作者备注:v4: substantially revised; new title, one added author. Adds a expanded 424-graphic benchmark, generalist-agent baselines, calibrated confidence scores, and downstream tasks: condition ranking and product prediction trained on a 60,103-reaction photochemistry database. Results in v1-v3 are superseded. 92 pages, 5 figures