Multi-class Topic Classification on AG News
89.4Title to Desc Alignment ScoreIMO-BART (Our)
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| IMO-BART (Our)Backbone=BART, Search Strategy=Bottom-up layer-wise search2024.04 | 89.4 | — | — | 81.97 | 85.68 | |
| IMO-BART lastBackbone=BART, Masking Strategy=Only the last layer2024.04 | 88.22 | — | — | 80.05 | 84.13 | |
| CHATGPTModel=gpt-3.5-turbo, Evaluation Protocol=few-shot in-context learning2024.04 | 85.13 | — | — | 79.28 | 82.21 | |
| PDAProtocol=Prompt-based learning2024.04 | 83.61 | — | — | 75.96 | 79.79 | |
| PADABackbone=T52024.04 | 82.39 | — | — | 75.52 | 78.96 | |
| BERTBackbone=BERT2024.04 | 81.11 | — | — | 67.95 | 74.68 | |
| BERT-EDABackbone=BERT, Data Augmentation=EDA2024.04 | 80.52 | — | — | 72.1 | 76.58 | |
| BERT-UDABackbone=BERT, Data Augmentation=UDA2024.04 | 80.41 | — | — | 71.81 | 75.82 | |
| BARTBackbone=BART2024.04 | 80.12 | — | — | 71.22 | 75.96 | |
| BERT-PGBBackbone=BERT, Data Augmentation=PGB2024.04 | 78.53 | — | — | 73.51 | 76.02 | |
| ALPACA-7BBackbone=LLaMA-7B, Evaluation Protocol=few-shot in-context learning2024.04 | 72.1 | — | — | 70.88 | 71.49 | |
| IMO-BART B2TBackbone=BART, Search Strategy=Bottom-up layer-wise search, Search Variant=B2T2024.04 | 70.31 | — | — | 64.59 | 67.45 | |
| IMO-BART w/o sqBackbone=BART, Search Strategy=Simultaneous search2024.04 | 62.59 | — | — | 57.27 | 59.93 | |
| ALPACA-7B-LoRABackbone=LLaMA-7B, Evaluation Protocol=Fine-tuned with low-rank adaptation2024.04 | 56.17 | — | — | 49.44 | 52.81 |