bag100/action-atlas-smolvla
Action Atlas: SmolVLA sparse autoencoders and concepts Sparse autoencoders (SAEs) and identified concepts for SmolVLA, part of the Action Atlas release accompanying the paper Not All Features Are Created Equal: A Mechanistic Study of Vision-Language-Action Models. The interactive explorer is at https://action-atlas.com. Paper: https://huggingface.co/papers/2603.19233 Code: https://github.com/CWRU-AISM/action-atlas Project page: https://cwru-aism.github.io/vla-interp-page… See the full description on the dataset page: https://huggingface.co/datasets/bag100/action-atlas-smolvla.
| Author / organization | bag100 |
|---|---|
| License (as declared) | cc-by-4.0 |
| Downloads | 190 (as reported by Hugging Face) |
| Likes | 0 |
| Created | 22 June 2026 |
| Last updated at source | 21 October 2026 |
Popularity, last 7 days
190 now (-161 since 2026-10-16). Counts as reported by the source, snapshotted daily by Silytics.
Connected in the knowledge graph
Recent changes
- 21 Oct 2026 Description: Action Atlas: SmolVLA sparse autoencoders and concepts Sparse autoencoders (SAEs) and identified concepts for SmolVLA, part of the Action Atlas release accompanying the paper on cross-task activation injection in vision-language-action mode → Action Atlas: SmolVLA sparse autoencoders and concepts Sparse autoencoders (SAEs) and identified concepts for SmolVLA, part of the Action Atlas release accompanying the paper Not All Features Are Created Equal: A Mechanistic Study of Vision
- 21 Oct 2026 Topics added task_categories:robotics
Changes detected by Silytics between refreshes of the source.
Related papers
More from bag100
Silytics shows public metadata from Hugging Face and links to the original. Information may be incomplete or out of date. Last refreshed 22 October 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.