Speech Processing on SUPERB
0.988KWS AccSSLAM
Evaluation results
| Method | Links | ||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| SSLAMParams=88M, Category=General Audio SSL2026.06 | 0.988 | 0.426 | — | 0.046 | 0.626 | 0.516 | — | — | — | 27.8 | — | — | 0.564 | — | — | — | |
| SPEARs+a XLarge# Param=600M, Pre-train data=197k2025.10 | 0.9812 | 0.963 | 0.0286 | 0.02 | 0.733 | 0.996 | 0.1173 | — | 2.9 | 3.15 | 2.75 | — | — | — | — | — | |
| SPEAR XLargeParams=600M, Category=Multi-domain Audio Models2026.06 | 0.981 | 0.963 | — | 0.032 | 0.733 | 0.996 | — | — | — | 3.2 | — | — | 0.029 | — | — | — | |
| SPEARs+a Large# Param=327M, Pre-train data=97k2025.10 | 0.9792 | 0.952 | 0.033 | 0.023 | 0.717 | 0.994 | 0.116 | — | 3.1 | 3.39 | 2.73 | — | — | — | — | — | |
| WavLM LargeParams=317M, Category=Speech Models2026.06 | 0.979 | 0.955 | — | 0.032 | 0.706 | 0.993 | — | — | — | 3.4 | — | — | 0.031 | — | — | — | |
| SPEAR LargeParams=327M, Category=Multi-domain Audio Models2026.06 | 0.979 | 0.95 | — | 0.038 | 0.716 | 0.994 | — | — | — | 3.4 | — | — | 0.031 | — | — | — | |
| SPEARs Large# Param=327M, Pre-train data=84k2025.10 | 0.9788 | 0.956 | 0.0314 | 0.023 | 0.721 | 0.9944 | 0.1167 | — | 2.6 | 3.24 | 2.72 | — | — | — | — | — | |
| WavLM Large# Param=317M, Pre-train data=94k2025.10 | 0.9786 | 0.955 | 0.0377 | 0.032 | 0.706 | 0.9931 | 0.1119 | — | 3.1 | 3.44 | 2.7 | — | — | — | — | — | |
| Dasheng 1.2B# Param=1.2B, Pre-train data=272k2025.10 | 0.9773 | 0.924 | — | 0.038 | 0.687 | 0.9813 | — | — | 14.3 | 13.8 | — | — | — | — | — | — | |
| BEATs iter3Params=90M, Category=General Audio SSL2026.06 | 0.977 | 0.571 | — | 0.052 | 0.645 | 0.534 | — | — | — | 25.9 | — | — | 0.364 | — | — | — | |
| USAD 2.0 Large+Params=336M, Teacher Type=Supervised Teachers2026.06 | 0.977 | 0.901 | — | 0.036 | 0.717 | 0.991 | — | — | — | 5.1 | — | — | 0.043 | — | — | — | |
| SPEAR BaseParams=94M, Category=Multi-domain Audio Models2026.06 | 0.976 | 0.9 | — | 0.041 | 0.694 | 0.981 | — | — | — | 3.8 | — | — | 0.039 | — | — | — | |
| USAD 2.0 XLarge+Params=695M, Teacher Type=Supervised Teachers2026.06 | 0.976 | 0.916 | — | 0.036 | 0.725 | 0.99 | — | — | — | 5.7 | — | — | 0.041 | — | — | — | |
| SPEARs+a Base# Param=94M, Pre-train data=97k2025.10 | 0.9759 | 0.903 | 0.0385 | 0.03 | 0.692 | 0.9813 | 0.1139 | — | 3.8 | 3.79 | 2.67 | — | — | — | — | — | |
| SPEARs Base# Param=94M, Pre-train data=84k2025.10 | 0.975 | 0.905 | 0.0377 | 0.029 | 0.691 | 0.9917 | 0.1132 | — | 3.4 | 3.43 | 2.64 | — | — | — | — | — | |
| WavLM Base+Params=95M, Category=Speech Models2026.06 | 0.974 | 0.894 | — | 0.035 | 0.687 | 0.99 | — | — | — | 5.6 | — | — | 0.039 | — | — | — | |
| WavLM Base+# Param=95M, Pre-train data=94k2025.10 | 0.9737 | 0.894 | 0.0407 | 0.035 | 0.687 | 0.99 | 0.1085 | — | 3.5 | 3.92 | 2.63 | — | — | — | — | — | |
| OLIVE-A (Mix+Gain)Params=93.16M2026.06 | 0.973 | 0.63 | 0.055 | 0.043 | 0.678 | 0.989 | 0.88 | 16.8 | — | — | — | — | 0.052 | 0.061 | 0.0459 | 0.244 | |
| OLIVE-JParams=93.16M2026.06 | 0.973 | 0.831 | 0.058 | 0.046 | 0.646 | 0.981 | 0.88 | 15.4 | — | — | — | — | 0.063 | 0.064 | 0.0445 | 0.258 | |
| USAD Base# Param=94M, Pre-train data=126k2025.10 | 0.971 | 0.886 | — | 0.042 | 0.68 | 0.983 | — | — | 5.1 | 7.7 | — | — | — | — | — | — | |
| USAD Large# Param=330M, Pre-train data=126k2025.10 | 0.971 | 0.912 | — | 0.039 | 0.684 | 0.984 | — | — | 4 | 6.5 | — | — | — | — | — | — | |
| USAD BaseParams=97M, Category=Multi-domain Audio Models2026.06 | 0.971 | 0.886 | — | 0.042 | 0.68 | 0.983 | — | — | — | 7.7 | — | — | 0.051 | — | — | — | |
| USAD LargeParams=336M, Category=Multi-domain Audio Models2026.06 | 0.971 | 0.912 | — | 0.039 | 0.684 | 0.984 | — | — | — | 6.5 | — | — | 0.04 | — | — | — | |
| WavLM BaseParams=94.38M2026.06 | 0.97 | 0.623 | 0.058 | 0.054 | 0.649 | 0.985 | 0.89 | 16.3 | — | — | — | — | 0.056 | 0.062 | 0.0813 | 0.242 | |
| OLIVE-A (Mix)Params=93.16M2026.06 | 0.97 | 0.669 | 0.06 | 0.046 | 0.671 | 0.991 | 0.88 | 16.5 | — | — | — | — | 0.064 | 0.064 | 0.039 | 0.252 | |
| USAD 2.0 SmallParams=25M, Teacher Type=Self-supervised Teachers2026.06 | 0.969 | 0.725 | — | 0.054 | 0.671 | 0.956 | — | — | — | 9.7 | — | — | 0.075 | — | — | — | |
| USAD 2.0 LargeParams=336M, Teacher Type=Self-supervised Teachers2026.06 | 0.969 | 0.847 | — | 0.048 | 0.678 | 0.951 | — | — | — | 5.5 | — | — | 0.039 | — | — | — | |
| USAD SmallParams=25M, Category=Multi-domain Audio Models2026.06 | 0.968 | 0.735 | — | 0.049 | 0.663 | 0.955 | — | — | — | 9.5 | — | — | 0.078 | — | — | — | |
| data2vec 2.0 BaseParams=93.16M2026.06 | 0.968 | 0.455 | 0.055 | 0.063 | 0.639 | 0.99 | 0.89 | 17.8 | — | — | — | — | 0.046 | 0.048 | 0.0556 | 0.234 | |
| USAD 2.0 BaseParams=97M, Teacher Type=Self-supervised Teachers2026.06 | 0.967 | 0.836 | — | 0.048 | 0.682 | 0.967 | — | — | — | 8.3 | — | — | 0.053 | — | — | — | |
| TAS-VAEAdaptive Weighting=true2026.04 | 0.9662 | 0.2311 | 0.088 | 0.0993 | 0.5677 | 0.7174 | — | — | 19.7 | 15.4 | — | 0.743 | — | — | — | — | |
| HuBERT BaseParams=94.70M2026.06 | 0.964 | 0.673 | 0.056 | 0.062 | 0.651 | 0.982 | 0.88 | 15.8 | — | — | — | — | 0.058 | 0.064 | 0.0748 | 0.251 | |
| DAS-VAEAdaptive Weighting=true2026.04 | 0.9636 | 0.2242 | 0.0883 | 0.1068 | 0.6018 | 0.7622 | — | — | 19.66 | 14.86 | — | 0.751 | — | — | — | — | |
| USAD 2.0 XLargeParams=695M, Teacher Type=Self-supervised Teachers2026.06 | 0.962 | 0.903 | — | 0.043 | 0.691 | 0.936 | — | — | — | 5.6 | — | — | 0.036 | — | — | — | |
| wav2vec 2.0 BaseParams=95.04M2026.06 | 0.96 | 0.665 | 0.058 | 0.066 | 0.625 | 0.926 | 0.87 | 14.9 | — | — | — | — | 0.064 | 0.065 | 0.0634 | 0.265 | |
| data2vec BaseParams=93.16M2026.06 | 0.959 | 0.549 | 0.07 | 0.07 | 0.654 | 0.986 | 0.89 | 17.5 | — | — | — | — | 0.038 | 0.05 | 0.0609 | 0.234 | |
| S3-RouterOptimal Sparsity=task-specific (0.06 to 0.1)2022.11 | 0.957 | 0.7107 | 0.0679 | 0.0718 | 0.6213 | 0.926 | 0.8876 | 0.1901 | — | — | — | — | — | — | — | — | |
| Weight ftmode=standard weight finetuning2022.11 | 0.955 | 0.668 | 0.0724 | 0.0741 | 0.615 | 0.9148 | 0.881 | 0.1885 | — | — | — | — | — | — | — | — | |
| ATST FrameParams=86M, Category=General Audio SSL2026.06 | 0.951 | 0.698 | — | 0.047 | 0.644 | 0.854 | — | — | — | 18.8 | — | — | 0.204 | — | — | — | |
| EATParams=88M, Category=General Audio SSL2026.06 | 0.928 | 0.45 | — | 0.047 | 0.625 | 0.537 | — | — | — | 25.9 | — | — | 0.55 | — | — | — | |
| JMAS-VAEAdaptive Weighting=true2026.04 | 0.9276 | 0.2458 | 0.0953 | 0.1065 | 0.5724 | 0.4848 | — | — | 36.72 | 21.04 | — | 0.681 | — | — | — | — | |
| MuQParams=330M, Category=Music SSL2026.06 | 0.915 | 0.497 | — | 0.051 | 0.623 | 0.574 | — | — | — | 29.7 | — | — | 0.399 | — | — | — | |
| DAS-VAEAdaptive Weighting=false2026.04 | 0.8046 | 0.1992 | 0.1061 | 0.1254 | 0.5373 | 0.2855 | — | — | 52.84 | 27.83 | — | 0.599 | — | — | — | — | |
| TAS-VAEAdaptive Weighting=false2026.04 | 0.6456 | 0.1765 | 0.1167 | 0.1364 | 0.4461 | 0.1693 | — | — | 70.12 | 39.97 | — | 0.51 | — | — | — | — | |
| EAT# Param=88M, Pre-train data=5k2025.10 | 0.625 | 0.45 | — | 0.047 | 0.625 | 0.928 | — | — | 55 | 25.9 | — | — | — | — | — | — | |
| JMAS-VAEAdaptive Weighting=false2026.04 | 0.6157 | 0.1761 | 0.1195 | 0.1281 | 0.4618 | 0.1811 | — | — | 70.95 | 37.49 | — | 0.513 | — | — | — | — | |
| BEATs# Param=90M, Pre-train data=5k2025.10 | 0.534 | 0.571 | — | 0.052 | 0.645 | 0.977 | — | — | 36.4 | 36.4 | — | — | — | — | — | — | |
| EnCodecAdaptive Weighting=false2026.04 | 0.4846 | 0.1808 | 0.1205 | 0.141 | 0.5041 | 0.1004 | — | — | 80.75 | 28.9 | — | 0.489 | — | — | — | — | |
| Semantic-VAEAdaptive Weighting=false2026.04 | 0.4349 | 0.1597 | 0.1241 | 0.1521 | 0.4599 | 0.0925 | — | — | 85.71 | 41.75 | — | 0.449 | — | — | — | — | |
| Vanilla VAEAdaptive Weighting=false2026.04 | 0.298 | 0.0774 | 0.1464 | 0.1711 | 0.3687 | 0.0598 | — | — | 89.4 | 53.48 | — | 0.382 | — | — | — | — | |
| Baseline (Mel/Fbank)Adaptive Weighting=false2026.04 | 0.0863 | 0 | 0.0956 | 0.1005 | 0.3539 | 0.091 | — | — | 82.01 | 20.86 | — | 0.413 | — | — | — | — | |
| FBANK2026.06 | 0.083 | 0.001 | 0.107 | 0.113 | 0.284 | 0.046 | 0.63 | 2.6 | — | — | — | — | 0.821 | 0.231 | 0.0043 | 0.595 |