AntonKorznikov/feature_stories

Feature Stories Matched contrastive stories for difference-in-means concept vectors, released with From Stories to Features: Scaling Concept Vectors for AI Alignment & Interpretability. Each feature is a concept–antagonist pair, not a concept label alone. The only human text for a pair is a one- or two-sentence narrative_guidance. The generator then writes two stories from one shared setup, so topic, cast, and scene stay fixed while the target pole changes. One row is one… See the full description on the dataset page: https://huggingface.co/datasets/AntonKorznikov/feature_stories.

<iframe src="https://silytics.si/embed/item/a3fc5eff-359a-451b-a8be-1baa56fba704" width="420" height="190" style="border:0" loading="lazy" title="Silytics entry"></iframe>

Silytics shows public metadata from Hugging Face and links to the original. Information may be incomplete or out of date. Last refreshed 18 October 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.