Text Classification on AG News (Accuracy, F1)
94.5AccuracyDeBERTa
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| DeBERTa2026.04 | 94.5 | 94.48 | |
| BERT2026.04 | 94.48 | 94.45 | |
| DistilBERT2026.04 | 92.76 | 94.78 | |
| ELECTRA2026.04 | 91.8 | 91.8 | |
| ANT2026.04 | 91 | — | |
| ADEK (Anchor count)=7002026.04 | 90.64 | 90.62 | |
| TOASTApprox.=4 → 5, Params=143.4M, Approximator=Linear, GFLOPs=22.2, tokens/s=1380.3, Backbone=ModernBERT-B2024.10 | 88 | — | |
| ModernBERT-BApprox.=original, Params=149.01M, Approximator=Linear, GFLOPs=23.25, tokens/s=1337.25, Backbone=ModernBERT-B2024.10 | 88 | — | |
| ModernBERT-BApprox.=original, Params=149.01M, Approximator=MLP, GFLOPs=23.25, tokens/s=1347.46, Backbone=ModernBERT-B2024.10 | 88 | — | |
| TOASTApprox.=11 → 12, Params=143.4M, Approximator=Linear, GFLOPs=22.2, tokens/s=1378.8, Backbone=ModernBERT-B2024.10 | 87 | — | |
| TOASTApprox.=20 → 21, Params=143.4M, Approximator=Linear, GFLOPs=22.2, tokens/s=1340.2, Backbone=ModernBERT-B2024.10 | 87 | — | |
| TOASTApprox.=4 → 8, Params=126.54M, Approximator=Linear, GFLOPs=19.0, tokens/s=1636.0, Backbone=ModernBERT-B2024.10 | 86 | — | |
| TOASTApprox.=11 → 14, Params=132.16M, Approximator=Linear, GFLOPs=20.1, tokens/s=1544.3, Backbone=ModernBERT-B2024.10 | 86 | — | |
| TOASTApprox.=18 → 21, Params=132.16M, Approximator=Linear, GFLOPs=20.1, tokens/s=1472.8, Backbone=ModernBERT-B2024.10 | 85 | — | |
| TOASTApprox.=20 → 21, Params=143.4M, Approximator=MLP, GFLOPs=22.20, tokens/s=1332.27, Backbone=ModernBERT-B2024.10 | 84 | — | |
| TOASTApprox.=4 → 8, 11 → 14, 18 → 21, Params=92.82M, Approximator=Linear, GFLOPs=12.7, tokens/s=2220.7, Backbone=ModernBERT-B2024.10 | 82 | — | |
| TOASTApprox.=4 → 7, 18 → 21, Params=109.68M, Approximator=Linear, GFLOPs=15.9, tokens/s=1803.9, Backbone=ModernBERT-B2024.10 | 82 | — | |
| TOASTApprox.=4 → 8, Params=126.54M, Approximator=MLP, GFLOPs=19.03, tokens/s=1632.65, Backbone=ModernBERT-B2024.10 | 82 | — | |
| TOASTApprox.=11 → 14, Params=132.16M, Approximator=MLP, GFLOPs=20.08, tokens/s=1540.23, Backbone=ModernBERT-B2024.10 | 82 | — | |
| TOASTApprox.=18 → 21, Params=132.16M, Approximator=MLP, GFLOPs=20.08, tokens/s=1467.56, Backbone=ModernBERT-B2024.10 | 82 | — | |
| TOASTApprox.=11 → 12, Params=143.4M, Approximator=MLP, GFLOPs=22.20, tokens/s=1394.63, Backbone=ModernBERT-B2024.10 | 82 | — | |
| TOASTApprox.=11 → 21, Params=92.82M, Approximator=Linear, GFLOPs=12.7, tokens/s=2264.0, Backbone=ModernBERT-B2024.10 | 81 | — | |
| TOASTApprox.=4 → 5, Params=143.4M, Approximator=MLP, GFLOPs=22.20, tokens/s=1384.42, Backbone=ModernBERT-B2024.10 | 81 | — | |
| TOASTApprox.=11 → 21, Params=92.82M, Approximator=MLP, GFLOPs=12.68, tokens/s=2216.50, Backbone=ModernBERT-B2024.10 | 73 | — | |
| TOASTApprox.=4 → 8, 11 → 14, 18 → 21, Params=92.82M, Approximator=MLP, GFLOPs=12.68, tokens/s=2155.16, Backbone=ModernBERT-B2024.10 | 73 | — | |
| TOASTApprox.=4 → 7, 18 → 21, Params=109.68M, Approximator=MLP, GFLOPs=15.85, tokens/s=1771.80, Backbone=ModernBERT-B2024.10 | 71 | — |