link GitHub
Browse files
README.md
CHANGED
|
@@ -17,3 +17,5 @@ Small artifacts from **Refusal is not Recognition** (Lexsi AI Safety & Alignment
|
|
| 17 |
| `jlens_earth_100.pt`, `jlens_earth_300.pt` | Jacobian lens (anthropics/jacobian-lens) fitted on Earth, on 100 and 300 wikitext prompts |
|
| 18 |
|
| 19 |
Abliterated weights are deliberately **not** released; the refusal direction plus SafeTune `orthogonalize_weights` reproduces them for research.
|
|
|
|
|
|
|
|
|
| 17 |
| `jlens_earth_100.pt`, `jlens_earth_300.pt` | Jacobian lens (anthropics/jacobian-lens) fitted on Earth, on 100 and 300 wikitext prompts |
|
| 18 |
|
| 19 |
Abliterated weights are deliberately **not** released; the refusal direction plus SafeTune `orthogonalize_weights` reproduces them for research.
|
| 20 |
+
|
| 21 |
+
Code, notebooks and full write-up: https://github.com/saiadith/refusal-vs-recognition
|