A 27B "AI Scientist" That Beats GPT-5.5 at Paper Replication
New work introduces Replica, a scalable task space for paper replication, and Faraday, a 27B-parameter agent that outperforms Claude Opus 4.8 and GPT-5.5 on held-out replication tasks. Replication, the authors argue, is the real proxy for scientific creativity and intuition, and better systems here edge closer to AI-enabled recursive self-improvement.