Extract and Evolve: How Claude Code and Complementary RL Architect Agent Memory
TL;DR: Through the lens of Claude Code’s memory mechanism, this post explores the philosophy of restraint and "subtractive" design in agent memory. We unpack the underlying logic behind our Complementary RL framework, analyze its architectural parallels with Claude, and share key lessons on evolving memory system offers a fresh perspective on LLM agents.
TL;DR: Let’s have a quick loot at how to involve FSDP2 into ROLL RL framework!
TL;DR: How does Layer-Norm scaling perform under different settings?
ScoreRS: Quality-Driven Curation of Remote Sensing Vision-Language Data via Learned Scoring Models
TL;DR: In this blog, I walk through the chain-of-thought behind my design for the ScoreRS project. I explain why I believe this work is important, detail the hypothesis and solutions I took to make it successful, and share the key challenges I overcome during development.