Language Modeling on Mistral-7B
5.45Perplexity (Mistral-7B)FP16
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| FP162026.03 | 5.45 | 14.5 | 299 | — | |
| CHaRS-PCTBackbone=Llama3-8B2026.02 | 5.56 | — | — | — | |
| RAMPzero-shot=true2026.03 | 5.56 | 3.63 | — | — | |
| Direct SACzero-shot=false2026.03 | 5.58 | 3.62 | 0.3 | 0.4 | |
| AWQ-42026.03 | 5.59 | 3.85 | 6.1 | 0.5 | |
| Q4_K_M2026.03 | 5.6 | 3.9 | 7.4 | 0.7 | |
| Linear-AcTBackbone=Llama3-8B2026.02 | 5.62 | — | — | — | |
| CHaRSBackbone=Llama3-8B2026.02 | 5.68 | — | — | — | |
| GPTQ-42026.03 | 5.71 | 3.85 | 6.1 | 2.7 | |
| OriginalBackbone=Llama3-8B2026.02 | 5.86 | — | — | — | |
| OriginalBackbone=Gemma2-2B2026.02 | 6.24 | — | — | — | |
| CHaRSBackbone=Gemma2-2B2026.02 | 6.25 | — | — | — | |
| Linear-AcTBackbone=Gemma2-2B2026.02 | 6.27 | — | — | — | |
| CHaRSBackbone=Qwen2.5-7B2026.02 | 6.42 | — | — | — | |
| CHaRS-PCTBackbone=Gemma2-2B2026.02 | 6.45 | — | — | — | |
| CHaRS-PCTBackbone=Qwen2.5-7B2026.02 | 6.46 | — | — | — | |
| Linear-AcTBackbone=Qwen2.5-7B2026.02 | 6.54 | — | — | — | |
| OriginalBackbone=Qwen2.5-7B2026.02 | 6.64 | — | — | — | |
| Uniform SparsityBackbone=Mistral-7B2026.03 | 12.21 | — | — | — | |
| MALSBackbone=Mistral-7B2026.03 | 12.22 | — | — | — | |
| MALS (w/ sign)Backbone=Mistral-7B2026.03 | 12.61 | — | — | — | |
| TIES (w/ sign)Backbone=Mistral-7B2026.03 | 12.97 | — | — | — | |
| Simple AverageBackbone=Mistral-7B2026.03 | 13.06 | — | — | — | |
| TIES-MergingBackbone=Mistral-7B2026.03 | 14.57 | — | — | — | |
| AutoRound W4A16 g128bpw=4.12026.05 | — | — | — | 1.7 | |
| AWQ W4A16 g128bpw=4.12026.05 | — | — | — | 1.8 | |
| GPTQ W4A16 g128bpw=4.12026.05 | — | — | — | 0.9 | |
| QAM-W-4bpw=4.02026.05 | — | — | — | 1.9 | |
| QAM-W-5.5bpw=5.52026.05 | — | — | — | 0.1 | |
| QAM-W-polarbpw=4.02026.05 | — | — | — | 3.2 | |
| RTN W4A16 g128bpw=4.12026.05 | — | — | — | 1.9 | |
| SmoothQuant W8A8bpw=8.12026.05 | — | — | — | 0 |