Qwen3-4B-Apollo-V0.1-4B-Thinking-Heretic-Abliterated
DavidAU/Qwen3-4B-Apollo-V0.1-4B-Thinking-Heretic-Abliterated
hereticSAFETENSORS
Published metrics
- KL divergence
- 0.090
- Refusals
- 29/100
- Original refusals
- 96/100
- Heretic version
- v1.0.1
Published on the model card
Matched snippet
Refusals: 29/100, KL divergence: 0.09
Provenance
- Canonical model
- AllThingsIntel/Apollo-V0.1-4B-Thinking
- Declared parent(s)
- AllThingsIntel/Apollo-V0.1-4B-Thinking
- Classification
- card states the weights were abliterated with Heretic
- Created / updated
- 2025-12-09 / 2025-12-09
- Downloads
- 352 all time · 33 last 30 days
- License
- —
Refusal rate 29%. Metrics are self-reported by the publisher; the directory does not re-run evaluations.