DeepSeek and Peking University introduce Engram, an O(1) conditional memory module that scales LLM sparsity beyond MoE
In tests, it boosted NIAH benchmark accuracy to 97%.
719028513.1K
Sentiment
Some users praised a profile picture tied to the DeepSeek Engram Module announcement while others accused the ideas of being stolen from OpenAI and labeled participants as low-quality trolls.
Pos
33.3%
Neg
66.7%
3 comments with sentiment.
Cluster Engagement
Digg Deeper
No Digg Deeper questions have been answered for this story yet.
Related links
Posts from X
Most Activity
Most Activity
VIEWS1.3KLIKES31RETWEETS1REPLIES1
Jack Morris@jxmnop
elie@eliebakouch
@jxmnop nice profile picture :o
1hViews 488Likes 13
/@gazorp5
@eliebakouch there's already a "applied ai research lab" founded in 2024 too
1hViews 165Likes 1
vyomakesh@juscallmevyom
@eliebakouch 💀
2hViews 189Likes 2
WilhelmH@lemmeyap
@eliebakouch Given it’s a deepseek paper the idea is probably stolen from OpenAI
1hViews 87
JMB 🧙♂️@jmbollenbacher
@eliebakouch @jxmnop Yoda continues to be a low quality troll poster.
Just mute him tbh.
1hViews 13
elie@eliebakouch
@gazorp5 i don’t think they ship much aha
58mViews 6