Language Modeling on WikiText-2 (PPL and Calibration Time)
4.91Perplexity (PPL)Base
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| BaseW/A=16/16, Model=LLaMA-2-13B2026.05 | 4.91 | — | |
| AWQW/A=4/16, Model=LLaMA-2-13B2026.05 | 5 | 49.6 | |
| AWQW/A=4/8, Model=LLaMA-2-13B2026.05 | 5 | 49.6 | |
| ORPW/A=4/16, Model=LLaMA-2-13B2026.05 | 5.09 | 34.8 | |
| ORPW/A=4/8, Model=LLaMA-2-13B2026.05 | 5.1 | 34.8 | |
| ORPW/A=3/16, Model=LLaMA-2-13B2026.05 | 5.32 | 34.8 | |
| ORPW/A=3/8, Model=LLaMA-2-13B2026.05 | 5.34 | 34.8 | |
| AWQW/A=3/16, Model=LLaMA-2-13B2026.05 | 5.47 | 49.6 | |
| AWQW/A=3/8, Model=LLaMA-2-13B2026.05 | 5.47 | 49.6 | |
| BaseW/A=16/16, Model=LLaMA-2-7B2026.05 | 5.49 | — | |
| AWQW/A=4/16, Model=LLaMA-2-7B2026.05 | 5.66 | 32.1 | |
| AWQW/A=4/8, Model=LLaMA-2-7B2026.05 | 5.66 | 32.1 | |
| SpinQuantW/A=4/8, Model=LLaMA-2-7B2026.05 | 5.7 | 358 | |
| ORPW/A=4/16, Model=LLaMA-2-7B2026.05 | 5.78 | 15.1 | |
| ORPW/A=4/8, Model=LLaMA-2-7B2026.05 | 5.79 | 15.1 | |
| ORPW/A=3/16, Model=LLaMA-2-7B2026.05 | 6.1 | 15.1 | |
| ORPW/A=3/8, Model=LLaMA-2-7B2026.05 | 6.11 | 15.1 | |
| AWQW/A=3/16, Model=LLaMA-2-7B2026.05 | 6.49 | 32.1 | |
| AWQW/A=3/8, Model=LLaMA-2-7B2026.05 | 6.5 | 32.1 |