Infrastructure for capturing LLM activations and SAE (Sparse Autoencoders) features, training probes for prompt maliciousness detection, and evaluating out-of-distribution generalization with Leave-One-Dataset-Out (LODO)
sparse-autoencoders out-of-distribution activation-analysis mechanistic-interpretability llm-security prompt-injection-detection activation-probing lodo-evaluation
-
Updated
Feb 18, 2026 - Python