Arjun Kocher
Independent AI Researcher (Lambda). I study deep neural architectures, and their underlying mathematics.
Blogs
-
Why Attention Is All You Need
A brief introduction to the seminal Transformer paper.
98370.6k -
GRPO
A groundbreaking training approach from DeepSeek.
1k62.3k -
Muon
A Brand New Optimizer from Kimi Moonshot.
34921.1k -
Chain of Thought
How reasoning emerges in large language models.
46029.4k -
LoZA
LongCat ZigZag Attention.
33516.8k -
XSA
Exclusive Self Attention.
42519.2k -
Nested Learning
A new learning paradigm.
1.5k92.9k -
HOPE
Architecture behind Nested Learning.
36222.2k -
DDL
Deep Delta Learning.
47031k -
RLM
Recursive Language Models.
1.6k67.6 -
mHC
Understanding Manifold Constrained Hyper Connections.
1.8k95.4k -
mHC math
Deriving Manifold Constrained Hyper Connections.
68334.5 -
Engram
A Memory technique introduced by DeepSeek.
30616k -
Attention Residuals
Understanding and Implementation.
1.6k247.4k
PyTorch Implementations
Notes
