Argument validity and novelty prediction on VALNOV (test)
45.16Joint F1ST-1st (GPT-3)
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| ST-1st (GPT-3)Category=Shared Task (ST)2023.05 | 45.16 | 74.64 | 61.32 | |
| CCKGFull System=true2023.05 | 43.91 | 70.69 | 63.3 | |
| ST-2nd (w/ KG)Category=Shared Task (ST)2023.05 | 43.27 | 69.8 | 62.43 | |
| contrastiveCategory=EG w/ Ours (ExplaGraphs)2023.05 | 37.82 | 64.77 | 59.96 | |
| T5Category=EG w/ Ours (ExplaGraphs)2023.05 | 37.71 | 67.07 | 63.53 | |
| max-marginCategory=EG w/ Ours (ExplaGraphs)2023.05 | 36.22 | 67.61 | 63.27 | |
| ST System AvgCategory=Shared Task (ST)2023.05 | 35.94 | 62.74 | 52.97 | |
| ST BaselineCategory=Shared Task (ST)2023.05 | 23.9 | 59.96 | 36.12 | |
| CCKGAblation=w/o PC-distance feats., Type=Relative to CCKG2023.05 | -2.27 | -0.27 | -3.73 | |
| CCKGAblation=w/o EWA, Type=Relative to CCKG2023.05 | -3.25 | -4.45 | 1.76 | |
| CCKGAblation=w/o connectivity feats., Type=Relative to CCKG2023.05 | -5.6 | -4.01 | -0.6 | |
| CCKGAblation=w/o EWO, Type=Relative to CCKG2023.05 | -6.51 | -3.8 | -1.69 | |
| CCKGAblation=string matching, Type=Relative to CCKG2023.05 | -6.71 | -3.23 | 0.55 | |
| CCKGAblation=w/o Graph feats., Type=Relative to CCKG2023.05 | -11.65 | -3.4 | -5.12 | |
| CCKGAblation=w/o Text feats., Type=Relative to CCKG2023.05 | -20.65 | -20.74 | -17.69 |