Story Cloze Test on ROCStories (test)
86.5AccuracyFTLM
Evaluation results
| Method | Links | |
|---|---|---|
| FTLMDescription=Requires large-scale pretrain2018.12 | 86.5 | |
| Our Model (All)Components=Content Unit + Logic Unit2018.12 | 79.1 | |
| HCMDescription=Relies on feature engineering2018.12 | 77.6 | |
| LS-SkipDescription=Fully neural network based2018.12 | 76.5 | |
| Logic Unit (LU)Ablation=Logic Unit only2018.12 | 75.6 | |
| HEPADescription=Fully neural network based2018.12 | 74.7 | |
| Content Unit (CU)Ablation=Content Unit only2018.12 | 71.7 |