Image Classification on CIFAR-10 (test) (Accuracy and Gap)
92.5AccuracyFP32 VGG Baseline
Evaluation results
| Method | Links | ||
|---|---|---|---|
| FP32 VGG BaselinePrecision=32-bit, Hardware=GPU2026.03 | 92.5 | — | |
| OursPrecision=4-bit, Hardware=CPU2026.03 | 92.34 | 0.16 | |
| 8-bit QATPrecision=8-bit, Hardware=GPU2026.03 | 90 | 1.5 | |
| DoReFa-NetPrecision=4-bit, Hardware=GPU2026.03 | 85 | 4.5 | |
| 4-bit PTQPrecision=4-bit, Hardware=GPU2026.03 | 70 | 14.5 |