paper
referenced-only
paper:sae-derivedSAE-derived indicators provide a sharp lens into per-sample distribution shift within LLMs, allowing the manifold-informed selection of samples for improved fine-tuning performance
Similar preprints — Semantic Scholar
Cited by (1)
- Understanding and Steering Llama 3 with Sparse Autoencoders - Goodfire
A sparse autoencoder trained on layer-19 residual-stream activations of Llama-3-8B-Instruct, using LMSYS-Chat-1M, yields features that can be labeled automatically, ranked by causal relevance, and use