General Language Evaluation on 5 Datasets Zero-shot
72.87Average AccuracyQuEPT
Evaluation Results
| Method | Links | |
|---|---|---|
| QuEPTModel=LLaMA-3-8B, Bits=W8A8, Zero-shot=true2026.02 | 72.87 | |
| FP16Model=LLaMA-3-8B, Bits=FP16, Zero-shot=true2026.02 | 72.65 | |
| QuEPTModel=LLaMA-3-8B, Bits=W6A6, Zero-shot=true2026.02 | 71.56 | |
| QuEPTModel=LLaMA-3-8B, Bits=W5A5, Zero-shot=true2026.02 | 70.99 | |
| SmoothQModel=LLaMA-3-8B, Bits=W6A6, Zero-shot=true2026.02 | 70.5 | |
| OmniQModel=LLaMA-3-8B, Bits=W6A6, Zero-shot=true2026.02 | 69.51 | |
| FP16Model=LLaMA-2-13B, Bits=FP16, Zero-shot=true2026.02 | 69.35 | |
| AffineQModel=LLaMA-3-8B, Bits=W6A6, Zero-shot=true2026.02 | 69.22 | |
| QuEPTModel=LLaMA-2-13B, Bits=W8A8, Zero-shot=true2026.02 | 69.09 | |
| QuEPTModel=LLaMA-2-13B, Bits=W6A6, Zero-shot=true2026.02 | 68.72 | |
| QuEPTModel=LLaMA-2-13B, Bits=W5A5, Zero-shot=true2026.02 | 68.39 | |
| QuEPTModel=LLaMA-3-8B, Bits=W4A4, Zero-shot=true2026.02 | 67.04 | |
| QuEPTModel=LLaMA-2-7B, Bits=W8A8, Zero-shot=true2026.02 | 66.23 | |
| QuEPTModel=LLaMA-2-13B, Bits=W4A4, Zero-shot=true2026.02 | 66.05 | |
| FP16Model=LLaMA-2-7B, Bits=FP16, Zero-shot=true2026.02 | 65.71 | |
| QuEPTModel=LLaMA-2-7B, Bits=W6A6, Zero-shot=true2026.02 | 65.48 | |
| DuQuantModel=LLaMA-3-8B, Bits=W4A4, Zero-shot=true2026.02 | 65.05 | |
| OmniQModel=LLaMA-2-13B, Bits=W6A6, Zero-shot=true2026.02 | 64.71 | |
| QuEPTModel=LLaMA-2-7B, Bits=W5A5, Zero-shot=true2026.02 | 64.42 | |
| SmoothQModel=LLaMA-2-13B, Bits=W6A6, Zero-shot=true2026.02 | 64.3 | |
| DuQuantModel=LLaMA-2-13B, Bits=W4A4, Zero-shot=true2026.02 | 62.84 | |
| QuEPTModel=LLaMA-2-7B, Bits=W4A4, Zero-shot=true2026.02 | 61.62 | |
| SmoothQModel=LLaMA-2-7B, Bits=W6A6, Zero-shot=true2026.02 | 61.4 | |
| DuQuantModel=LLaMA-2-7B, Bits=W4A4, Zero-shot=true2026.02 | 59.11 | |
| OmniQModel=LLaMA-2-7B, Bits=W6A6, Zero-shot=true2026.02 | 58.56 | |
| QuaRotModel=LLaMA-3-8B, Bits=W4A4, Zero-shot=true2026.02 | 57.16 | |
| QuaRotModel=LLaMA-2-7B, Bits=W4A4, Zero-shot=true2026.02 | 54.55 | |
| QLLMModel=LLaMA-2-13B, Bits=W4A4, Zero-shot=true2026.02 | 54.31 | |
| OmniQModel=LLaMA-2-13B, Bits=W4A4, Zero-shot=true2026.02 | 53.13 | |
| QLLMModel=LLaMA-2-7B, Bits=W4A4, Zero-shot=true2026.02 | 51.6 | |
| OmniQModel=LLaMA-2-7B, Bits=W4A4, Zero-shot=true2026.02 | 49.13 | |
| OmniQModel=LLaMA-3-8B, Bits=W4A4, Zero-shot=true2026.02 | 35.7 | |
| AffineQModel=LLaMA-3-8B, Bits=W4A4, Zero-shot=true2026.02 | 35.44 |