Blog

In new research, we find that the CoT is informative about LLM cognition as long as the cognition is complex enough that it can’t be performed in a single forward pass

Backpropogatable FDFD and FDTD simulators in Jax, and trying to use diffusion models, tiled preconditioning, and other tricks to try to make them faster

Research on redundant attention heads in language models and their role in in-context learning through Bayesian updates.

A look at how language models update their priors based on in-context examples