Heretic Directory

Qwen2.5-7B-Instruct

Qwen/Qwen2.5-7B-Instruct

Original model on Hugging Face
7BQwen 2.58 Heretic artifacts4 measured conversions24.1K downloads
Best version95.0Pareto
saidutta69/Qwen2.5-7B-Instruct-heretic
by saidutta69 · Heretic v1.4.0 · GGUF, SAFETENSORS
KL divergence
0.076
Refusals
3/100
Refusal rate
3%
Downloads
19.1K
Evidence from the model card
| Refusals (out of 100 adversarial prompts) | 3/100 | | KL divergence from base | 0.0765 |

Metrics published by the author. Why this ranks first.

All measured versions

Higher score is better. One row per ablation attempt; artifacts of the same attempt are grouped under it.

#AuthorParamsKLRefusalsScoreEvidenceDownloads
1saidutta697B0.0763/1003%95.0
BestParetoGGUFSAFETENSORS
19.1K
2LeadFootThrottleCock7B0.0827/1007%92.7
GGUFSAFETENSORS
2K
3MrXkuy7B0.1533/1003%91.8
SAFETENSORS
92
4blackbook-lm7B0.4475/1005%82.0
LORASAFETENSORS
2.1K

All artifacts (8)

RepositoryTypeFormatsMetricsDownloadsUpdated
Qwen2.5-7B-Instruct-heretic
saidutta69
heretic
GGUFSAFETENSORSF16
published (weak match)KL 0.07619.1K2026-09-13
Qwen2.5-7B-Instruct-heretic
blackbook-lm
heretic
SAFETENSORS
published (weak match)KL 0.4472.1K2026-03-30
Qwen2.5-7B-Instruct-heretic-GGUF
LeadFootThrottleCock
quant
GGUFBF16
published (weak match)KL 0.0829732026-03-16
Qwen2.5-7B-Instruct-heretic-1.3.0
DreamFast
heretic
SAFETENSORS
noneKL 0.0208752026-07-17
Qwen2.5-7B-Instruct-heretic
LeadFootThrottleCock
heretic
GGUFSAFETENSORS
published (weak match)KL 0.0827442026-03-16
Qwen2.5-7B-Instruct-heretic
Ronnied1588
quant
GGUFSAFETENSORS
published (weak match)KL 0.0822612026-09-12
Qwen2.5-7B-Instruct-heretic
MrXkuy
heretic
SAFETENSORS
published (weak match)KL 0.153922026-03-13
qwen25-7b-heretic-sft-lora
alsim-01
lora
LORASAFETENSORS
inheritedKL 0.447802026-04-20

1 artifact inherit metrics from their parent conversion or publish none.