Blog
Thoughts on engineering, ML, and building products.
August 31, 2026
Random In vs. Random Out: Evaluating Initial State Variance against Token-Wise Sampling in Open-Weight LLMs
Empirical benchmark comparing Latent Space Perturbation ('Random In') vs. Token-Level Temperature Sampling ('Random Out') across reasoning domains in LLMs to evaluate output diversity and representation collapse.
deep learningmachine learningLLM
August 6, 2026
What AI Actually Does in Math Research
I am just an undergraduate student. Here is what AI actually helps with, and where it completely falls short.
aimath
July 17, 2026
I Derived a New Attention Formula (Cosine-RoPE Attention) and Tested It. Here's What Happened
Cosine-RoPE Attention formula that shows almost similar result.
deep learningmachine learning