Multi-task Classification on GLUE (MRPC, RTE, CoLA) (test, val)
88.95MRPC AccuracyDSGD
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| DSGDBackbone=Roberta-large, Multi-task Setting=MRPC + CoLA2026.04 | 88.95 | — | 85.21 | — | — | |
| PCGradBackbone=Roberta-large, Multi-task Setting=MRPC + RTE2026.04 | 88.63 | 83.25 | — | — | — | |
| DSGDBackbone=Roberta-large, Multi-task Setting=MRPC + RTE2026.04 | 88.14 | 81.73 | — | — | — | |
| LNSRBackbone=Roberta-large, Multi-task Setting=MRPC + CoLA2026.04 | 88.01 | — | 83.74 | — | — | |
| FFTBackbone=Roberta-large, Multi-task Setting=MRPC + RTE2026.04 | 85.22 | 76.03 | — | — | — | |
| NoisyTuneBackbone=Roberta-large, Multi-task Setting=MRPC + RTE2026.04 | 84.78 | 75.02 | — | — | — | |
| LNSRBackbone=Roberta-large, Multi-task Setting=MRPC + RTE2026.04 | 81.18 | 69.75 | — | — | — | |
| SWABackbone=Roberta-large, Multi-task Setting=MRPC + RTE2026.04 | 79.68 | 65.56 | — | — | — | |
| SWABackbone=Roberta-large, Multi-task Setting=MRPC + CoLA2026.04 | 79.63 | — | 79.01 | — | — | |
| FFTBackbone=Roberta-large, Multi-task Setting=MRPC + CoLA2026.04 | 77.33 | — | 77.3 | — | — | |
| PCGradBackbone=Roberta-large, Multi-task Setting=MRPC + CoLA2026.04 | 76.67 | — | 75.66 | — | — | |
| NoisyTuneBackbone=Roberta-large, Multi-task Setting=MRPC + CoLA2026.04 | 70.8 | — | 71.74 | — | — | |
| DSGDBackbone=Roberta-large, Multi-task Setting=RTE + CoLA2026.04 | — | 83.1 | 85.35 | — | — | |
| DSGDBackbone=Roberta-large, Multi-task Setting=Aggregate2026.04 | — | — | — | 85.41 | 1.19 | |
| FFTBackbone=Roberta-large, Multi-task Setting=RTE + CoLA2026.04 | — | 68.3 | 80.51 | — | — | |
| FFTBackbone=Roberta-large, Multi-task Setting=Aggregate2026.04 | — | — | — | 77.44 | 10.75 | |
| LNSRBackbone=Roberta-large, Multi-task Setting=RTE + CoLA2026.04 | — | 79.24 | 83.59 | — | — | |
| LNSRBackbone=Roberta-large, Multi-task Setting=Aggregate2026.04 | — | — | — | 80.91 | 9.08 | |
| NoisyTuneBackbone=Roberta-large, Multi-task Setting=RTE + CoLA2026.04 | — | 68.16 | 81.14 | — | — | |
| NoisyTuneBackbone=Roberta-large, Multi-task Setting=Aggregate2026.04 | — | — | — | 75.27 | 9.26 | |
| PCGradBackbone=Roberta-large, Multi-task Setting=RTE + CoLA2026.04 | — | 80.94 | 83.99 | — | — | |
| PCGradBackbone=Roberta-large, Multi-task Setting=Aggregate2026.04 | — | — | — | 81.52 | 6.19 | |
| SWABackbone=Roberta-large, Multi-task Setting=RTE + CoLA2026.04 | — | 76.79 | 85.54 | — | — | |
| SWABackbone=Roberta-large, Multi-task Setting=Aggregate2026.04 | — | — | — | 77.68 | 9.59 |