New interpretability research just dropped by my friends Jonathan Jacobi and Gal Niv! They demonstrate a method to interpret the contributions of individual MLPs in LLMs to the residual stream, which were too subtle for existing methods like PatchScopes to interpret. Check it out:

Introducing Multiverse: the first AI-generated multiplayer game. Multiplayer was the missing piece in AI-generated worlds — now it’s here. Players can interact and shape a shared AI-simulated world, in real-time. Training and research cost < $1.5K. Run it on your own PC. We

Multiplayer: ✅

This is the coolest dataset I've seen on Hugging Face today: action labels of 1v1 races in Gran Turismo 4 to train a multiplayer world model! Great stuff by Enigma!