HiFC swaps LLM KV caches directly between GPU and pSLC-SSD, matching DRAM-level throughput while eliminating DRAM and slashing cost five-fold.

We propose Diffusion Adaptive Text Embedding (DATE), which improves text-to-image diffusion models by dynamically refining text embeddings throughout the diffusion sampling process.

Larger vocabulary lowers language modeling difficulty by facilitating models to learn non-i.i.d patterns in text more easily