Word Sense Disambiguation on CoarseWSD-20
90.6AccuracyZ-LaVI (SBERT)
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Z-LaVI (SBERT)Zero-shot=true, labeled pre-training=true, Visual Imagination=true2022.10 | 90.6 | 87.5 | |
| Z-LaVI (SimCSE)Zero-shot=true, Visual Imagination=true2022.10 | 89.7 | 86 | |
| BERT-L-3shotk-shot=3, labeled pre-training=true2022.10 | 89.3 | 85.2 | |
| Z-LaVI (GPT-Neo-2.7B)Zero-shot=true, Visual Imagination=true2022.10 | 88.9 | 85.3 | |
| Z-LaVI (OPT-30B)Zero-shot=true, Visual Imagination=true2022.10 | 88.8 | 85.1 | |
| Z-LaVI (GPT-J-6B)Zero-shot=true, Visual Imagination=true2022.10 | 88.5 | 84.8 | |
| Z-LaVI (GPT-Neo-1.3B)Zero-shot=true, Visual Imagination=true2022.10 | 88.4 | 84.6 | |
| SBERTZero-shot=true, labeled pre-training=true2022.10 | 87.8 | 83.3 | |
| Z-LaVI w/o LMZero-shot=true, vision-text prediction only=true2022.10 | 87.7 | 83.8 | |
| GPT-Neo-2.7BZero-shot=true2022.10 | 86.7 | 81.5 | |
| SimCSEZero-shot=true2022.10 | 85.1 | 78.9 | |
| GPT-Neo-1.3BZero-shot=true2022.10 | 84.7 | 78.3 | |
| OPT-30BZero-shot=true2022.10 | 84.4 | 80.4 | |
| GPT-J-6BZero-shot=true2022.10 | 84.1 | 79.3 | |
| Z-LaVI (ROBERTa-L-mnli)Zero-shot=true, labeled pre-training=true, Visual Imagination=true2022.10 | 83 | 78.1 | |
| Z-LaVI (BART-L-mnli)Zero-shot=true, labeled pre-training=true, Visual Imagination=true2022.10 | 82.4 | 77.9 | |
| ROBERTa-L-mnliZero-shot=true, labeled pre-training=true2022.10 | 80.4 | 74.4 | |
| BART-L-mnliZero-shot=true, labeled pre-training=true2022.10 | 80.2 | 74.8 | |
| BERT-L-1shotk-shot=1, labeled pre-training=true2022.10 | 77.6 | 71.2 | |
| RandomZero-shot=true2022.10 | 41.3 | 36.7 |