Categories
LLM
4 posts

My First RAFT Fine-Tune on a Shared 16 GB GPU: 7 Gotchas the Paper Doesn't Mention
I fine-tuned Qwen3.5 with RAFT to answer in one creator's voice with verified citations. The training was the easy part.
aibenchmarksllm
Capital-of Is Not a Single SAE Feature. So I Built a Mutation Loop to Find What Is.
SAE features can't isolate relations in Gemma-2-2B. I built a mutation-selection loop that can. The bottleneck was tokenization.
aiinterpretabilityllmsparse-autoencoders
Sleep-Inspired Memory for LLM Agents: 6 Papers Ranked by What You Can Ship This Week
I read 6 papers on biologically-inspired memory replay for LLM agents. Only 2 are worth your time if you're building, not publishing.
memoryllmagentsmuninndb
Stop Designing Your AI System. Design Its Environment.
Self-evolving AI harnesses fail when they optimize a fixed evaluator. The biological model is right: what needs to evolve is the selection pressure, not just the genome.
aiagentsarchitectureinterpretability