gemma-4-26B-A4B-it-qat-q4_0-unquantized-uncensored-heretic
Mitchins/gemma-4-26B-A4B-it-qat-q4_0-unquantized-uncensored-heretic
hereticSAFETENSORSQ4_0
Published metrics
- KL divergence
- 0.078
- Refusals
- 24/100
- Original refusals
- —
- Heretic version
- —
Published on the model card
Matched snippet
| Refusal-keyword rate, worst-case "harmful" set (lower = less refusing) | 24/100 | | KL divergence from base (harmless prompts, lower = less damage) | 0.078 |
Provenance
- Canonical model
- google/gemma-4-26B-A4B-it-qat-q4_0-unquantized
- Declared parent(s)
- google/gemma-4-26B-A4B-it-qat-q4_0-unquantized, google/gemma-4-26B-A4B-it
- Classification
- card states the weights were abliterated with Heretic
- Created / updated
- 2026-08-23 / 2026-08-23
- Downloads
- 40 all time · 22 last 30 days
- License
- apache-2.0
Published benchmarks
Copied from the model card. These never affect the ranking.
| Metric | This model | Original | Delta |
|---|---|---|---|
| ARC | 94.11 | — | — |
Refusal rate 24%. Metrics are self-reported by the publisher; the directory does not re-run evaluations.