Arjun Kocher — Independent AI Researcher

Writing

• 0
  Why Attention Is All You Need

  A brief introduction to the seminal Transformer paper.
  983
• 0
  GRPO

  A groundbreaking training approach from DeepSeek.
  1k
• 0
  Muon

  A Brand New Optimizer from Kimi Moonshot.
  349
• 0
  Chain of Thought

  How reasoning emerges in large language models.
  460
• 0
  LoZA

  LongCat ZigZag Attention.
  335
• 0
  XSA

  Exclusive Self Attention.
  425
• 0
  Nested Learning

  A new learning paradigm.
  1.5k
• 0
  HOPE

  Architecture behind Nested Learning.
  362
• 0
  DDL

  Deep Delta Learning.
  470
• 0
  RLM

  Recursive Language Models.
  1.6k
• 0
  mHC

  Understanding Manifold Constrained Hyper Connections.
  1.8k
• 0
  mHC math

  Deriving Manifold Constrained Hyper Connections.
  683
• 0
  Compressed Sparse Attention

  A Faithful Implementation of CSA from DeepSeek V4.
  502
• 0
  Engram

  A Memory technique introduced by DeepSeek.
  306
• 0
  DeepSeek V4

  Technical Report rundown of DeepSeek V4.
  833
• 0
  Attention Residuals

  Understanding and Implementation.
  1.6k
• 0
  RL Algorithm Questions

  RL interview questions 2026
  1.7k
• 0
  Kimi Delta Attention

  The Introduction of Kimi K3 Architecture from Kimi Moonshot.
  870