Our work isn’t done. We’re currently working on incorporating search and risk of bias assessments. If you’re interested in connecting, please don’t hesitate to reach out!
12.06.2025 17:05
👍 2
🔁 0
💬 0
📌 0
Our work isn’t done. We’re currently working on incorporating search and risk of bias assessments. If you’re interested in connecting, please don’t hesitate to reach out!
... Bijan Teja; Alexander Leung; Lina Ghosn; Rahul Arora; Michael Noetel; David B. Emerson
Plus all other researchers involved: Katherine Manta; Elina Farahani; Matthew Cecere; Anabel Selemon; Linsey Gong; Robert Kloosterman; Scott Jiang; Richard Saleh; Denis Margalik; James Lin; Jane Jomy; Jerry Xie; David Chen; Jaswanth Gorla; Sylvia Lee; Kelvin Zhang; Harriet Ware; Mairead Whelan...
A special shout out to:
@rohitarorayyc.bsky.social
@paullusla.bsky.social
@jzsang.bsky.social
for working tirelessly to make this dream happen <3
And our advisors:
@isabelleboutron.bsky.social
David Moher
@geochurch.bsky.social
@nikbobrovitz.bsky.social
Check out our blog post: ottosr.com/blog/announc...
Preprint: l.ottosr.com/preprint
If you’re a researcher who works on systematic reviews: we’re offering otto-SR through a free pilot at ottosr.com/sign-up/
In one review, otto-SR found 5 more relevant papers than the original authors (9 vs 4 studies).
These additional studies revealed that taking immune-enhancing nutrition before gastric surgery likely reduces hospital stay by over 1 day (p<0.05).
When we reproduced an entire Cochrane issue of reviews (n=12 reviews), otto-SR found nearly twice as many relevant studies (n=114) compared to the original reviews (n=64).
These additional studies materially changed conclusions: two reviews became statistically significant, one lost significance.
Through rigorous benchmarking, otto-SR consistently outperformed or matched humans in:
- screening sensitivity (otto-SR: 96.7% vs. human: 81.7%)
- screening specificity (otto-SR 97.9% vs. human 98.1%)
- data extraction accuracy (otto-SR: 93.1% vs. human: 79.7%)
Today we’re announcing otto-SR, an AI workflow to perform systematic reviews 3000x faster
By using gpt-4.1 and o3-mini, ottoSR beats humans at all tasks
In two days, we conducted 12 work-years of Cochrane research–finding 2x more relevant papers and altering key conclusions 🧵
