Content
@
https://warpcast.com/~/channel/technology
0 reply
0 recast
0 reaction
glb
@glb
Transformers have dominated LLM text generation, and generate tokens sequentially. This is a cool attempt to explore diffusion models as an alternative, by generating the entire text at the same time using a coarse-to-fine process. Congrats @StefanoErmon & team!
0 reply
0 recast
0 reaction
bridgitmendler
@bridgitmendler
I’ve been working on a hybrid LLM since 2023, combining transformers (like GPT) with GANs in an integrated architecture. The toughest part was getting GANs to work with discrete data, but I played around with techniques like Gumbel-Softmax and policy gradients to handle the
0 reply
0 recast
0 reaction
