Natural Language Inference on ChaosNLI S_amb (test)
41.8Top-1 AccuracyModel A (projection target)
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Model A (projection target)Train section=train_S_easy, Val section=val_S_easy, Learning Rate=0.22026.04 | 41.8 | 0.035 | 0.011 | |
| Model C (vote-proportion target)Train section=train_S_easy, Val section=val_full, Learning Rate=0.22026.04 | 41.5 | — | — | |
| Model A (projection target)Train section=train_S_amb, Val section=val_S_amb, Learning Rate=0.0092026.04 | 41 | 0.037 | 0.022 | |
| Model C (vote-proportion target)Train section=train_S_easy, Val section=val_S_easy, Learning Rate=0.32026.04 | 40.6 | — | — | |
| Model C (vote-proportion target)Train section=train_S_amb, Val section=val_full, Learning Rate=0.0032026.04 | 40.5 | — | — | |
| Model A (projection target)Train section=train_S_amb, Val section=val_S_easy, Learning Rate=0.012026.04 | 40.4 | 0.032 | 0.047 | |
| Model A (projection target)Train section=train_S_amb, Val section=val_full, Learning Rate=0.0092026.04 | 40.1 | 0.01 | 0.004 | |
| Model A (projection target)Train section=train_full, Val section=val_S_amb, Learning Rate=0.52026.04 | 40 | 0.009 | 0.027 | |
| Model A (projection target)Train section=train_S_easy, Val section=val_S_amb, Learning Rate=0.082026.04 | 39.5 | 0.024 | 0.014 | |
| Model B (antipignistic target)Train section=train_full, Val section=val_S_amb, Learning Rate=0.62026.04 | 39.1 | — | — | |
| Model B (antipignistic target)Train section=train_S_amb, Val section=val_full, Learning Rate=0.62026.04 | 39.1 | — | — | |
| Model C (vote-proportion target)Train section=train_S_amb, Val section=val_S_amb, Learning Rate=0.72026.04 | 38.9 | — | — | |
| Model B (antipignistic target)Train section=train_S_easy, Val section=val_S_easy, Learning Rate=0.092026.04 | 38.2 | — | — | |
| Model C (vote-proportion target)Train section=train_S_easy, Val section=val_S_amb, Learning Rate=0.32026.04 | 38.1 | — | — | |
| Model B (antipignistic target)Train section=train_S_easy, Val section=val_full, Learning Rate=0.12026.04 | 38 | — | — | |
| Model C (vote-proportion target)Train section=train_full, Val section=val_S_amb, Learning Rate=0.00012026.04 | 37.3 | — | — | |
| Model B (antipignistic target)Train section=train_S_amb, Val section=val_S_amb, Learning Rate=0.72026.04 | 37.3 | — | — | |
| Model B (antipignistic target)Train section=train_S_amb, Val section=val_S_easy, Learning Rate=0.82026.04 | 37.2 | — | — | |
| Model B (antipignistic target)Train section=train_S_easy, Val section=val_S_amb, Learning Rate=0.72026.04 | 37.1 | — | — | |
| Model C (vote-proportion target)Train section=train_full, Val section=val_full, Learning Rate=0.0052026.04 | 35.8 | — | — | |
| Model C (vote-proportion target)Train section=train_S_amb, Val section=val_S_easy, Learning Rate=0.52026.04 | 35.7 | — | — | |
| Model A (projection target)Train section=train_S_easy, Val section=val_full, Learning Rate=0.0052026.04 | 34.7 | 0.033 | 0.068 | |
| Model B (antipignistic target)Train section=train_full, Val section=val_full, Learning Rate=0.0072026.04 | 34.3 | — | — | |
| Model A (projection target)Train section=train_full, Val section=val_S_easy, Learning Rate=0.032026.04 | 34.1 | 0 | 0.009 | |
| Model B (antipignistic target)Train section=train_full, Val section=val_S_easy, Learning Rate=0.0072026.04 | 34.1 | — | — | |
| Model A (projection target)Train section=train_full, Val section=val_full, Learning Rate=0.0072026.04 | 33.7 | 0.006 | 0.022 | |
| Model C (vote-proportion target)Train section=train_full, Val section=val_S_easy, Learning Rate=0.082026.04 | 33.2 | — | — |