Llama-3.1-Hawkish-8B
mukaj/Llama-3.1-Hawkish-8B
8BLlama 3.13 Heretic artifacts1 measured conversions12.5K downloads
Best version93.2Pareto
ChiKoi7/Llama-3.1-Hawkish-8B-Hereticby ChiKoi7 · Heretic v1.1.0 · GGUF, SAFETENSORS
- KL divergence
- 0.059
- Refusals
- 8/100
- Refusal rate
- 8%
- Downloads
- 12.5K
Evidence from the model card
| Refusals | 8/100 | | KL divergence | 0.0595 |
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.
| # | Author | Params | KL | Refusals | Score | Evidence | Downloads |
|---|---|---|---|---|---|---|---|
| 1 | ChiKoi7 | 8B | 0.059 | 8/1008% | 93.2 | BestParetoGGUFSAFETENSORS | 12.5K |
All artifacts (3)
| Repository | Type | Formats | Metrics | Downloads | Updated |
|---|---|---|---|---|---|
| Llama-3.1-Hawkish-8B-Heretic-i1-GGUF mradermacher | quant | GGUFIQ1_M | inheritedKL 0.059 | 8.9K | 2025-12-23 |
| Llama-3.1-Hawkish-8B-Heretic-GGUF mradermacher | quant | GGUFIQ4_XS | inheritedKL 0.059 | 3.5K | 2025-12-23 |
| Llama-3.1-Hawkish-8B-Heretic ChiKoi7 | heretic | SAFETENSORS | published (weak match)KL 0.059 | 128 | 2025-12-23 |