paper
referenced-only
paper:robustly-identifying-e434e9Robustly identifying concepts introduced during chat fine-tuning using crosscoders
Similar preprints — Semantic Scholar
Cited by (2)
- Understanding sparse autoencoder scaling in the presence of feature manifolds
Sparse autoencoder scaling laws can break down when activations contain feature manifolds: if per-feature loss falls slowly with latent count, an SAE spends its latents tiling common manifolds instead
- Persona Features Control Emergent Misalignment
Fine-tuning GPT-4o on narrowly incorrect datasets—spanning nine domains including health, legal, and automotive advice—generalizes into broadly malicious behavior across unrelated prompts, confirming