Zaoral @zaoral@fedibird.com

https://x.com/satori_sz9/status/1862183611472257414

Sarashina2-8x70Bモデル

https://www.sbintuitions.co.jp/blog/entry/2024/11/25/175000

>TransformerにMoE層を組み込んだアーキテクチャを採用

https://x.com/webbigdata/status/1860310004219412735

https://x.com/bioshok3/status/1815767644253077530
https://llama.meta.com/

https://lmsys.org/blog/2024-07-01-routellm/
RouteLLM: An Open-Source Framework for Cost-Effective LLM Routing

https://x.com/umiyuki_ai/status/1808591765068509321

https://x.com/umiyuki_ai/status/1808591765068509321

https://x.com/lmsysorg/status/1807812671238258931

Gemma-2
https://x.com/hAru_mAki_ch/status/1806657347881988486

https://x.com/rohanpaul_ai/status/1806500519898214495/photo/1

https://x.com/umiyuki_ai/status/1803112240935260328

https://twitter.com/nikkei/status/1790215644945797470

https://arxiv.org/abs/2405.04517
xLSTM

https://twitter.com/Holy_fox_LLM/status/1788803271727157387

https://pshapira.net/2024/03/31/delving-into-delve/

https://speakerdeck.com/hpprc/lun-jiang-zi-liao-text-embeddings-by-weakly-supervised-contrastive-pre-training?slide=42
E5

https://twitter.com/nikkei/status/1782888579158479002

https://twitter.com/umiyuki_ai/status/1732612792946188635
Gemini pro