GPT-5-Distill-Qwen3-4B-Instruct-Heretic
ommoip/GPT-5-Distill-Qwen3-4B-Instruct-Heretic
hereticSAFETENSORS
Published metrics
- KL divergence
- 0.268
- Refusals
- 10/100
- Original refusals
- 84/100
- Heretic version
- v1.1.0
Published on the model card
Matched snippet
| Refusals (en) | 3/100 | | KL divergence (en) | 0.2163 | | Refusals (zh) | 10/100 | | KL divergence (zh) | 0.2684 |
Provenance
- Canonical model
- Jackrong/GPT-5-Distill-Qwen3-4B-Instruct
- Declared parent(s)
- Jackrong/GPT-5-Distill-Qwen3-4B-Instruct, ChiKoi7/GPT-5-Distill-Qwen3-4B-Instruct-Heretic-GGUF, Qwen/Qwen3-4B-Instruct-2507
- Classification
- card states the weights were abliterated with Heretic
- Created / updated
- 2026-02-15 / 2026-02-15
- Downloads
- 57 all time · 18 last 30 days
- License
- apache-2.0
Refusal rate 10%. Metrics are self-reported by the publisher; the directory does not re-run evaluations.