Blog

Writing

Blog

Notes on reinforcement learning, generative models, and research observations — written to think through ideas, not to summarize papers.

What's happening in my field →
2025

Test-Time Inverse Scaling in Audio LLMs

 ·  reinforcement learning · audio LLMs · reasoning · test-time scaling

Chain-of-thought reasoning helps text LLMs but hurts Audio LLMs. This post explains why — and how process rewards fix it.

Read →

The Exploration-Exploitation Dilemma in RLHF for Generative Models

 ·  reinforcement learning · RLHF · generative models · flow matching

A deep dive into why fixed regularization in RLHF leads to diversity collapse, and how adaptive sample-level control resolves the exploration-exploitation dilemma.

Read →