Guiding LLM Post-training Data Engineering with Model Internals from Sparse Autoencoders Paper • 2605.27354 • Published May 26 • 15
