Rosetta Activations

Rosetta Activations Updated: 2026-06-15 02:30 UTC Contrastive activation extractions for 17 semantic concepts across 46 language models, supporting cross-architecture mechanistic interpretability research. Companion concept pair corpus: jamesrahenry/Rosetta Concept Pairs Papers: forthcoming Dataset Structure Rosetta-Activations/ ├── rcp v1/ Current extraction line — richest data (N≈2000) │ └── {Model Name}/ │ ├── calibration {concept}.npy… See the full description on the dataset page:

Source: Hugging Face Hub (james-ra-henry/Rosetta-Activations). Metadata imported from the dataset’s Hub tags.