Popular repositories Loading

1. lora-X-Interpretability lora-X-Interpretability Public

   We are investigating what LoRA fine-tuning does to a language model's internal feature representations using Sparse Autoencoders — specifically whether adapters reuse existing base model features o…