Templeton and colleagues: scaled feature dictionaries for Claude 3 Sonnet

What changed

Scaled sparse-autoencoder dictionaries to a production language model, investigating abstract multilingual and multimodal features and feature-based behavioral interventions.

Scientific scope

Scaled sparse-autoencoder dictionaries to a production language model, investigating abstract multilingual and multimodal features and feature-based behavioral interventions. Only selected internal activations and imperfect learned dictionaries are studied. Features can split or be absent; this is not an exhaustive map or demonstrated general safety control.

Evidence

The linked original sources describe the stated contribution and its evidence limits. WRS has not independently rerun the experiments or certified a proof. Only selected internal activations and imperfect learned dictionaries are studied. Features can split or be absent; this is not an exhaustive map or demonstrated general safety control.

Contributors

• Source-listed authors and technical contributors · SCIENCE: Source-listed authors or explicitly credited technical contributors for the cited versions. Family entries retain source-specific attribution; a combined list is not one original-paper byline or equal individual work. Names alone do not establish resolved identities.

• Scaling Monosemanticity: Extracting Interpretable Features from Claude 3 Sonnet
• Templeton and colleagues: scaled feature dictionaries for Claude 3 Sonnet — official research record

Provisional WRS assessment, edition 2026-10-16.1. Independent expert calibration is pending. This grade does not certify correctness or assign ownership.

Full scientific record · Browse contribution grades · Request a correction