AI1 min read
Base Labs joined Hugging Face and Goodfire AI to build safety tools for open models. They aim to make safety transparent and part of the training process.
Why it matters: Developers need safer open models because removing safeguards creates dangerous versions called abliterated models.
By OpenSmartRoute editorial · written through the router by llm-onprem
From TechCrunch AI - “Base Labs launches an open-weight AI safety partnership with Hugging Face and Goodfire”
Base Labs launched a new partnership on Wednesday with Hugging Face and Goodfire AI.
The group will build infrastructure for evaluating and monitoring open-weight models.
This comes after reports that over 6,000 models have had their safeguards removed.
Base Labs plans to create standards for training and monitoring these systems.
They want safety built into the model lifecycle rather than added later.
Goodfire specializes in explaining how AI models make decisions internally.
Keep reading
AI1 min read
OpenAI released a new framework to track and disclose misalignment issues in its models. It includes three review tracks and six detailed incident reports from reinforcement learning training.
AI1 min read
Nunchux AI launched VC-Attention to speed up video diffusion models without retraining. It uses low-bit quantization and replaces the slow softmax stage.
AI1 min read
Knowledgator released GLiFormer, an encoder that extracts nested JSON without generating tokens.
