angkul’s site

blogs

notes

projects

daily shelf


Topological Signatures of Deception: Detecting Unfaithful Reasoning via Sentence-Level Causal Graphs

A Mechanistic Switch for Heuristic vs. Algorithmic Processing in Small Language Models

https://angkul.bearblog.dev/the-surgical-beauty-of-non-linear-ffn/

Scaling Laws for Neural Language Models

Sliding Window Attention Training paper(SWAT)

What is Knowledge Graph and GraphRAG