ResearchAI-written article
Trillium Labs Bets Open Research Can Make High-Risk AI Easier to Scrutinize
The new nonprofit plans to publish experiments on post-training, reinforcement learning and recursive self-improvement so outside researchers can examine and replicate the work.