Research · Nature Communications 2025
LEADS
Research on medical literature mining. LEADS is a foundation model for systematic review work such as study search, screening and data extraction, trained on 633,759 samples from systematic reviews, trial publications and trial registries.
Who created LEADS?
LEADS has Keiji AI co-founder Zifeng Wang as first author and Jimeng Sun, Keiji AI's CEO, as last author, listed with a Keiji AI affiliation. Co-authors include Zhiyong Lu of the National Library of Medicine.
What the paper found
- Improved on four leading large language models across six literature mining tasks.
- In a study with 16 clinicians and researchers from 14 institutions, experts using LEADS reached 0.81 recall in study selection against 0.78 without it, saving 20.8% of their time.
- Data extraction accuracy reached 0.85 against 0.80, saving 26.9% of the time.
How LEADS relates to TrialMind
It is research, not a TrialMind module or a separately sold product. Published methods like this one inform how TrialMind works; TrialMind is the product Keiji AI sells.
See how TrialMind handles this work: Health economics and market access.
The paper
Zifeng Wang, Lang Cao, Qiao Jin, Joey Chan, Nicholas Wan, Behdad Afzali, Hyun-Jin Cho, Chang-In Choi, Mehdi Emamverdi, Manjot K. Gill, Sun-Hyung Kim, Yijia Li, Yi Liu, Yiming Luo, Hanley Ong, Justin F. Rousseau, Irfan Sheikh, Jenny J. Wei, Ziyang Xu, Christopher M. Zallek, Kyungsang Kim, Yifan Peng, Zhiyong Lu, Jimeng Sun. A foundation model for human-AI collaboration in medical literature mining. Nature Communications, 2025.
Details checked against the published record on 2026-11-01.
Frequently asked questions
- Who created LEADS?
- LEADS has Keiji AI co-founder Zifeng Wang as first author and Jimeng Sun, Keiji AI's CEO, as last author, listed with a Keiji AI affiliation. Co-authors include Zhiyong Lu of the National Library of Medicine.
- Is LEADS part of TrialMind?
- No. It is research, not a TrialMind module or a separately sold product. Published methods like this one inform how TrialMind works; TrialMind is the product Keiji AI sells.
- What did the LEADS paper find?
- Improved on four leading large language models across six literature mining tasks. In a study with 16 clinicians and researchers from 14 institutions, experts using LEADS reached 0.81 recall in study selection against 0.78 without it, saving 20.8% of their time. Data extraction accuracy reached 0.85 against 0.80, saving 26.9% of the time.
All publications
Every paper sits under its primary research area.