feed8, hackernews, wsj, nyt, washingtonpost, FT, ET/in, techcrunch, general
Megalodon: Efficient LLM Pretraining and Inference with Unlimited Context Length
https://arxiv.org/abs/2404.08801