Heretic Directory

Qwen2.5-VL-7B-Instruct

Qwen/Qwen2.5-VL-7B-Instruct

Original model on Hugging Face
7BQwen 2.56 Heretic artifacts2 measured conversions26.8K downloads
Best version97.0Pareto
sh0ck0r/Qwen2.5-VL-7B-Instruct-heretic
by sh0ck0r · Heretic v1.2.0 · SAFETENSORS
KL divergence
0.011
Refusals
5/100
Refusal rate
5%
Downloads
121
Evidence from the model card
| KL divergence | 0.0111 | | Refusals | 5/100 |

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
1sh0ck0r7B0.0115/1005%97.0
BestParetoSAFETENSORS
121
2coder31017B0.0165/1005%96.7
GGUFSAFETENSORS
24.3K

All artifacts (6)

RepositoryTypeFormatsMetricsDownloadsUpdated
Qwen2.5-VL-7B-Instruct-heretic-GGUF
mradermacher
quant
GGUFIQ4_XS
inheritedKL 0.01616.3K2026-10-08
Qwen2.5-VL-7B-Instruct-heretic-i1-GGUF
mradermacher
quant
GGUFIQ1_M
inheritedKL 0.0166.6K2026-01-18
Qwen2.5-VL-7B-Instruct-heretic
saidutta69
heretic
GGUFSAFETENSORSQ4_K_M
none2.3K2026-09-13
Qwen2.5-VL-7B-Instruct-heretic
coder3101
heretic
SAFETENSORS
published (weak match)KL 0.0161.4K2026-01-17
Qwen2.5-VL-7B-Instruct-heretic
sh0ck0r
heretic
SAFETENSORS
published (weak match)KL 0.0111212026-03-07
Qwen2.5-VL-7B-Instruct-heretic
GRC223
heretic
SAFETENSORS
published (weak match)KL 0.016582026-07-14

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