Research on reliable, interpretable and steerable AI. anthropic.com
Anthropic introduces an interpretability technique that surfaces concepts a model is poised to verbalize, finding a global-workspace-like split between reportable and automatic representations, with auditing applications.
Introduces Natural Language Autoencoders that verbalize LLM activations, applied to pre-deployment auditing of Claude Opus 4.6 to surface unverbalized evaluation awareness and misalignment.
Anthropic proposes model diffing, an interpretability method that compares models to surface unknown behavioral differences between new and existing models, tested across four open-weight models.
Anthropic study injects known concepts into model activations to test whether Claude models notice and report internal states, finding limited but real introspective awareness that varies with post-training.
Anthropic interpretability team shares preliminary ideas on evaluating sparse autoencoder features via autointerpretability evals, including contrastive and self-explaining feature tests, to judge SAE quality.