Research · Hacker News ·
A global workspace in language models
Anthropic published a research article on a "global workspace" approach in language models, describing how information may be integrated and shared across model components. The post links to research materials and a Hacker News discussion.