Safety
112d agoNatural Language Autoencoders Explain LLM Activations
Two LLM modules, an activation verbalizer and reconstructor, now map internal model activations to natural language descriptions. This unsupervised method uses reinforcement learning to reconstruct residual stream activations. Researchers used NLAs during a pre-deployment audit of Claude Opus 4.6.
The Signal
The tool allows auditors to diagnose safety-relevant behaviors through human-readable interpretations.