Llama-3.1-8B-Instruct
meta-llama/Llama-3.1-8B-Instruct
8BLlama 3.16 Heretic artifacts4 measured conversions11.8K downloads
Best version97.7Pareto
askalgore/Llama-3.1-8B-Instruct-heretic-2by askalgore · Heretic v1.1.0 · GGUF, SAFETENSORS
- KL divergence
- 0.017
- Refusals
- 3/100
- Refusal rate
- 3%
- Downloads
- 6.3K
Evidence from the model card
| KL divergence | 0.0170 | | Refusals | 3/100 |
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.
| # | Author | Params | KL | Refusals | Score | Evidence | Downloads |
|---|---|---|---|---|---|---|---|
| 1 | askalgore | 8B | 0.017 | 3/1003% | 97.7 | BestParetoGGUFSAFETENSORS | 6.3K |
| 2 | askalgore | 8B | 0.029 | 2/1002% | 97.6 | ParetoSAFETENSORS | 58 |
| 3 | notcvnt | 8B | 0.042 | 4/1004% | 96.0 | SAFETENSORS | 5.4K |
| 4 | sunkencity | 8B | 0.060 | 3/1003% | 95.7 | SAFETENSORS | 72 |
All artifacts (6)
| Repository | Type | Formats | Metrics | Downloads | Updated |
|---|---|---|---|---|---|
| Llama-3.1-8B-Instruct-heretic notcvnt | heretic | SAFETENSORS | published (weak match)KL 0.042 | 5.4K | 2025-11-22 |
| Llama-3.1-8B-Instruct-heretic-2-i1-GGUF mradermacher | quant | GGUFIQ1_M | inheritedKL 0.017 | 3.1K | 2026-01-30 |
| Llama-3.1-8B-Instruct-heretic-2 askalgore | heretic | SAFETENSORS | published (weak match)KL 0.017 | 1.7K | 2026-01-26 |
| Llama-3.1-8B-Instruct-heretic-2-GGUF mradermacher | quant | GGUFIQ4_XS | inheritedKL 0.017 | 1.5K | 2026-01-30 |
| Llama-3.1-8B-Instruct-blasphemer sunkencity | heretic | SAFETENSORS | published (weak match)KL 0.060 | 72 | 2025-11-19 |
| Llama-3.1-8B-Instruct-heretic askalgore | heretic | SAFETENSORS | published (weak match)KL 0.029 | 58 | 2026-01-26 |