Egor Dmitriev
Search
Search
Recent Posts
How to Break a Transformer's Loss Plateau (and Why Waiting Is a Lottery)
Oct 22, 2026
Is Half Your Context Window Still Marketing in 2026?
LLM Comprehension
Sep 26, 2026
Let Your Types Carry the Rules
Programming with Invariants
Sep 12, 2026
When Your Types Are Arrays
Programming with Invariants
Sep 12, 2026
All posts →
Tag: training-dynamics
1 item with this tag.
Oct 22, 2026
How to Break a Transformer's Loss Plateau (and Why Waiting Is a Lottery)
transformers
training-dynamics
in-context-learning
induction-heads
optimization