Parts-of-Speech as Emergent Categories in SAE Latent Space
Local Sparsity Enables Unsupervised LLM Safety Detection
SAEScientist-Bench: Can AI Agents Conduct Autonomous SAE Interpretability Research?
2
SAEScientist-Bench: Can AI Agents Conduct Autonomous SAE Interpretability Research?
1
LLM Forensics: Where Do Backdoors Hide? Localizing and Controlling Trigger Mechanisms with Sparse Autoencoders
1