Announcement
Goodfire says it will embed researchers at OpenAI to probe model internals
OpenAI says Goodfire will analyze model internals to help monitor training, debug models and improve alignment.
TLDR
Goodfire says its researchers will get deep access to OpenAI’s internal models and systems to monitor training runs, debug unwanted behavior and inform alignment improvements. The company calls this work “model forensics.” OpenAI researchers will be the primary audience for its findings, Goodfire says, and it also plans to publish learnings as the collaboration progresses.
Combined views
49.6K
19 Sources, first seen ago

Comments