Lightweight Guide to understanding GRPO and RL principles
A beginner-friendly guide to Group Relative Policy Optimization (GRPO) training workflow without assuming prior RL knowledge.
Recent posts
Bridging the Three Gulfs of Agentic Development (and how they shape evals)
A practical framework for spotting and fixing evaluation blind spots in agentic LLM pipelines, based on Shankar et al.’s Three Gulfs model.
Let Agents do the talking: A Scalable Way to Evaluate Multi-Turn Chatbots
Interactive evaluations: lightweight, automated tests that use agents to measure multi-turn chatbot quality at scale.
CUDA Study Log 4: Optimizing Constrained Decoding with Triton Kernel
Update traditional CUDA matrix multiplication kernel for constrained decoding
CUDA Studylog 3 - Tiling and Shared Memory for Matrix Multiplication Optimization
Optimizing CUDA matrix multiplication using tiling and shared memory, with detailed explanations of memory access patterns and performance improvements