Hugging FaceCapability change· 16 OctFixing Gradient AccumulationFixes a bug in the Transformers Trainer where gradient accumulation loss was not computed correctly for token-level tasks such as causal LM training.
Mistral AI· 16 OctUn Ministral, des MinistrauxThe source text is minimal, only including the title, date, and author of a blog post, with no substantive announcement details.
OpenAIStory of the week· 15 OctEvaluating fairness in ChatGPTOpenAI analyzed how ChatGPT responds to users based on their name, using AI research assistants to protect privacy.