A nonprofit research institute that releases fully open models such as OLMo.
Current rank (1M)
58
Rank over the last 23 days
2026.09.08 – 2026.09.30 · High 52 · Low 80 · Now 58
Google Research published work on September 17 on interactive learning simulations that teachers build themselves. A single prompt cleared every requirement only 3.5% of the time, and ten rounds of self-correction pushed that to 69.3%.
The AI's hypothesis was verified twice — once against a separate dataset, once against lab imaging — and the result led to an expanded partnership. Invasive lobular carcinoma, long considered resistant to immunotherapy, accounts for about 15% of US breast cancer diagnoses.
At an event marking an expanded partnership with Providence Swedish, Ai2 addressed where AI still falls short on judgment and verification in science
By cycling through evaluation, data synthesis, and retraining, it runs specialized models up to 100x cheaper than frontier models
In an evaluation built from real tutoring transcripts, LLMs struggled to tell when to help a student and when to hold back
The independent evaluator known for measuring model performance is now preparing a tool developers can use directly
Preview version measures the ability to distinguish between when to help a student and when to let them think for themselves
That's the last story.
