Zero-shot Commonsense Reasoning Suite Subset (ARC-c, BoolQ, PIQA, HellaSwag, WinoGrande)
55.03ARC-c AccuracyFP
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| FPModel=LLaMA-3 8B, #Bits=162026.05 | 55.03 | 83.05 | 81.12 | 79.3 | 74.58 | 74.62 | |
| WINDQuantModel=LLaMA-3 8B, #Bits=1.962026.05 | 49.9 | 77.1 | 78.4 | 75.4 | 70.3 | 70.22 | |
| FPModel=LLaMA-3 3B, #Bits=162026.05 | 46.3 | 74.1 | 78.2 | 74.1 | 69.5 | 67.61 | |
| AQLMModel=LLaMA-3 8B, #Bits=2.272026.05 | 45.7 | 75.6 | 78.4 | 73.9 | 71.6 | 69.04 | |
| WINDQuantModel=LLaMA-3 3B, #Bits=1.992026.05 | 37.8 | 66.8 | 73.9 | 65.7 | 65.7 | 61.98 | |
| AQLMModel=LLaMA-3 3B, #Bits=2.272026.05 | 37.6 | 70.5 | 73.8 | 64.4 | 63 | 61.86 | |
| FPModel=LLaMA-3 1B, #Bits=162026.05 | 36.7 | 63.8 | 74.8 | 64.2 | 60.7 | 60.04 | |
| SignRoundModel=LLaMA-3 8B, #Bits=22026.05 | 33.1 | 64.6 | 69.3 | 56.7 | 60.1 | 56.75 | |
| SignRoundModel=LLaMA-3 3B, #Bits=22026.05 | 31.3 | 61.5 | 67.4 | 48.9 | 56.5 | 53.13 | |
| EfficientQATModel=LLaMA-3 8B, #Bits=2.122026.05 | 31.1 | 60.6 | 70.3 | 56.3 | 59.3 | 55.52 | |
| WINDQuantModel=LLaMA-3 1B, #Bits=2.012026.05 | 28 | 52.7 | 64.3 | 47 | 54.5 | 49.3 | |
| SpinQuantModel=LLaMA-3 3B, #Bits=22026.05 | 27 | 46.2 | 52.4 | 26.4 | 47.9 | 39.98 | |
| OmniQModel=LLaMA-3 1B, #Bits=22026.05 | 26.2 | 37.8 | 53.6 | 25.9 | 48.9 | 38.48 | |
| AWQModel=LLaMA-3 1B, #Bits=22026.05 | 26 | 50.7 | 52.2 | 26 | 50.7 | 41.12 | |
| SignRoundModel=LLaMA-3 1B, #Bits=22026.05 | 26 | 53.2 | 59.8 | 33.3 | 53 | 45.09 | |
| SpinQuantModel=LLaMA-3 1B, #Bits=22026.05 | 25.7 | 44 | 50.7 | 26.5 | 51.5 | 39.68 | |
| LLM-QATModel=LLaMA-3 3B, #Bits=22026.05 | 25.7 | 37.8 | 50.7 | 41.5 | 54.4 | 42.02 | |
| AWQModel=LLaMA-3 8B, #Bits=22026.05 | 25.7 | 42.9 | 50.3 | 25.9 | 48.5 | 38.66 | |
| OmniQModel=LLaMA-3 8B, #Bits=22026.05 | 25.7 | 37.8 | 53.1 | 26.4 | 50.6 | 38.72 | |
| SpinQuantModel=LLaMA-3 8B, #Bits=22026.05 | 25.7 | 47.7 | 50.8 | 26.2 | 49.5 | 39.98 | |
| AWQModel=LLaMA-3 3B, #Bits=22026.05 | 25.5 | 37.8 | 52.9 | 26.3 | 50.2 | 38.54 | |
| LLM-QATModel=LLaMA-3 1B, #Bits=22026.05 | 25.2 | 37.8 | 52.3 | 25.6 | 49.1 | 38 | |
| GPTQModel=LLaMA-3 1B, #Bits=22026.05 | 25 | 41.2 | 50.6 | 26.9 | 51 | 38.94 | |
| EfficientQATModel=LLaMA-3 3B, #Bits=2.122026.05 | 24.8 | 43 | 64.1 | 37.1 | 55 | 44.8 | |
| TesseraQModel=LLaMA-3 3B, #Bits=22026.05 | 24.6 | 54.6 | 60.1 | 37.3 | 54.1 | 46.13 | |
| EfficientQATModel=LLaMA-3 1B, #Bits=2.122026.05 | 24.3 | 60.3 | 59.8 | 35.7 | 52.8 | 46.58 | |
| OmniQModel=LLaMA-3 3B, #Bits=22026.05 | 24.2 | 37.8 | 50.7 | 26.1 | 48.9 | 37.54 | |
| GPTQModel=LLaMA-3 8B, #Bits=22026.05 | 24.2 | 50.5 | 52.7 | 30.2 | 50.4 | 41.6 | |
| GPTQModel=LLaMA-3 3B, #Bits=22026.05 | 24 | 43.7 | 51.3 | 27.7 | 47.7 | 38.88 | |
| GPTVQModel=LLaMA-3 1B, #Bits=22026.05 | 23.8 | 45.9 | 55.7 | 31.2 | 50.6 | 41.44 | |
| TesseraQModel=LLaMA-3 1B, #Bits=22026.05 | 23.3 | 40.4 | 52.8 | 30 | 50.4 | 39.38 | |
| TesseraQModel=LLaMA-3 8B, #Bits=22026.05 | 22.1 | 57.1 | 53.3 | 54 | 58.4 | 48.97 | |
| GPTVQModel=LLaMA-3 8B, #Bits=22026.05 | 21.9 | 37.8 | 53.9 | 27.7 | 50.5 | 38.37 | |
| LLM-QATModel=LLaMA-3 8B, #Bits=22026.05 | 21.6 | 37.8 | 50.3 | 26.1 | 49.9 | 37.14 | |
| GPTVQModel=LLaMA-3 3B, #Bits=22026.05 | 20.7 | 37.9 | 52.4 | 27.9 | 49.8 | 37.74 |