Observed arrival · 2026-10-29
Unlocked Models: a field guide to refusal-edited AI
A bilingual guide cataloging refusal-removed open models, their licenses, deployment requirements, and defensive considerations.
- For
- AI security researchers and open-model deployers
- Worth noticing
- The site says all 40 repositories were resolved through the Hugging Face Hub API in a scan dated 2026-10-27.
Field notes
The page distinguishes direction-based ablation, automated “heretic” searches, uncensored fine-tunes, and multi-direction edits rather than treating every refusal-removed checkpoint as the same thing. It says uploader counts are self-reported unless backed by a paper, and notes reported tradeoffs including logic and coding performance and refusals that were useful. English and Korean navigation carries through the major sections, though the extracted view does not include the full 40-row table.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
One card from the complete issue