Common-sense QA on ARC Easy
66.12AccuracyOPT-2.7B FP16
Evaluation results
| Method | Links | ||
|---|---|---|---|
| OPT-2.7B FP16Model=OPT-2.7B, Precision=FP162023.08 | 66.12 | 7.41 | |
| TSLDModel=OPT-2.7B, Quantization=QAT-KD (tensor-wise)2023.08 | 59.39 | 8.12 | |
| LogitModel=OPT-2.7B, Quantization=QAT-KD (tensor-wise)2023.08 | 58.92 | 9.05 | |
| GT+LogitModel=OPT-2.7B, Quantization=QAT-KD (tensor-wise)2023.08 | 58.84 | 8.66 |