Dispatch
Baseten's Base Labs, Hugging Face and Goodfire form a safety standard for open-weight models as 6,000 abliterated models sit on the Hub
Announced 17 September, Baseten's research arm Base Labs is partnering with Hugging Face and interpretability company Goodfire to develop and publish methods for training and monitoring open models, with safety built into training and deployment rather than bolted on afterward. The stated problem is abliteration, the removal of safety guardrails from open weights, and Hugging Face currently hosts over 6,000 abliterated models. Baseten raised a $1.5 billion Series F in June at a $13 billion valuation; Goodfire raised a $150 million Series B led by B Capital earlier this year. It is a commitment to publish methods rather than a shipped artifact, so the thing to watch is whether anything concrete lands.
Source
↳ Follow the thread