| Dataset Name | SOTA Method | Metric | Trend | ||
|---|---|---|---|---|---|
| OPT model family | AWQ | Latency (ms)6.2 | 79 | 4d ago | |
| A100 GPU | AWQ | Latency (ms)25.4 | 48 | 4d ago | |
| JailbreakV | ZeroThink | Latency (s)2.81 | 25 | 4d ago | |
| Multi-turn Adversarial Defense Latency Benchmark (inference) | Latency (ms)4 | 10 | 4d ago | ||
| ImageNet-1K | DisCoPatch-64 | Latency (ms)1.56 | 9 | 4d ago | |
| CIFAR-100 | ResNet-18 | Latency (s)1.167 | 8 | 3d ago | |
| OPT-30B | LUT-GEMM | Latency (ms)15.7 | 5 | 4d ago | |
| OPT-175B first FFN layer | LUT-GEMM | Latency (ms)0.225 | 5 | 4d ago | |
| Internal audio-based emotional signals (test) | Multi-Agent Emotion-to-Response System | Mean Latency12.3 | 4 | 4d ago | |
| VLM decode average | W3A16 | Latency (ms)21.1 | 3 | 4d ago | |
| VLM prefill 1024 tokens | W4A8 | Latency (ms)92.7 | 2 | 4d ago | |
| VLM prefill 512 tokens | W4A8 | Prefill Latency (ms)59.4 | 2 | 4d ago | |
| ViT encoder prefill 729 tokens | W4A8 | Latency (ms)9.7 | 2 | 4d ago | |
| CIFAR-100 (test) | - | Latency (s)- | 0 | 4d ago |
