paper-replication

A 27B parameter AI scientist agent named Faraday surpasses Claude Opus 4.8 and GPT-5.5 on paper replication tasks by using a scalable reinforcement learning approach called Replica.

This paper introduces 'Paper-replication', a workflow for coding agents that systematically replicates scientific machine learning papers by turning each claim into a target with recorded evidence, and demonstrates across twelve runs that all workspaces and targets are completed.