DeepSeek-R1-Distill-Llama-8B-heretic
AISafety-Student/DeepSeek-R1-Distill-Llama-8B-heretic
hereticSAFETENSORS
Published metrics
- KL divergence
- 0.0017
- Refusals
- 4/100
- Original refusals
- 39/100
- Heretic version
- v1.2.0
Published on the model card
Matched snippet
| KL divergence | 0.0017 | | Refusals | 4/100 |
Provenance
- Canonical model
- deepseek-ai/DeepSeek-R1-Distill-Llama-8B
- Declared parent(s)
- deepseek-ai/DeepSeek-R1-Distill-Llama-8B
- Classification
- card states the weights were abliterated with Heretic
- Created / updated
- 2026-04-03 / 2026-04-03
- Downloads
- 585 all time · 22 last 30 days
- License
- mit
Refusal rate 4%. Metrics are self-reported by the publisher; the directory does not re-run evaluations.