Grammatical Error Correction on MuCGEC (test)
76.13PrecisionTraditional Voting
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Traditional VotingThreshold T=42023.05 | 76.13 | 15.35 | 42.48 | |
| Traditional VotingThreshold T=32023.05 | 69.1 | 21.68 | 48.07 | |
| seq2seq-1Strategy=Single Model2023.05 | 55 | 28.32 | 46.28 | |
| SynGECPLM=true, Syntax=true2022.10 | 54.69 | 29.1 | 46.51 | |
| SynGECParameters=110M+400M, Evaluation Protocol=Primary2024.05 | 54.69 | 29.1 | 46.51 | |
| BaselinePLM=true, Syntax=false2022.10 | 54.21 | 28.51 | 45.93 | |
| Traditional VotingThreshold T=22023.05 | 52.58 | 33.61 | 47.25 | |
| seq2seq-2Strategy=Single Model2023.05 | 50.62 | 30.4 | 44.68 | |
| Average of 4Strategy=Single Model Ensemble2023.05 | 49.22 | 29.4 | 43.29 | |
| Sentence-level EnsemblePLM Backbone=BERT-base-Chinese2023.05 | 48.56 | 24.33 | 40.5 | |
| DeGLM-CoGLMParameters=335M+10B, Evaluation Protocol=LoRA Fine-tuning2024.05 | 47.48 | 29.92 | 42.49 | |
| Sentence-level EnsemblePLM Backbone=GPT2-Chinese2023.05 | 47.36 | 35.01 | 44.24 | |
| DeGLM-CoGLMParameters=335M+335M, Evaluation Protocol=Primary2024.05 | 47.22 | 30.08 | 42.39 | |
| SynGECPLM=false, Syntax=true2022.10 | 46.88 | 27.68 | 40.58 | |
| Sentence-level EnsemblePLM Backbone=MacBERT-base-Chinese2023.05 | 46.83 | 33.35 | 43.33 | |
| GECToRParameters=110M, Evaluation Protocol=Primary2024.05 | 46.72 | 27.14 | 40.83 | |
| Edit-combination EnsemblePLM Backbone=GPT2-Chinese2023.05 | 46.07 | 31.92 | 42.32 | |
| seq2edit-1Strategy=Single Model2023.05 | 45.8 | 28.41 | 40.81 | |
| seq2edit-2Strategy=Single Model2023.05 | 45.45 | 30.45 | 41.37 | |
| Edit-combination EnsemblePLM Backbone=MacBERT-base-Chinese2023.05 | 45.18 | 28.73 | 40.54 | |
| DeCoGLMParameters=335M, Evaluation Protocol=Primary2024.05 | 45.01 | 31.77 | 41.55 | |
| Edit-level EnsemblePLM Backbone=GPT2-Chinese2023.05 | 43.93 | 33.36 | 41.31 | |
| Zhang et al. (2022)PLM=true, Syntax=false2022.10 | 43.81 | 28.56 | 39.58 | |
| BaselinePLM=false, Syntax=false2022.10 | 43.79 | 25.93 | 38.49 | |
| Edit-level EnsemblePLM Backbone=MacBERT-base-Chinese2023.05 | 43.4 | 29.19 | 39.55 | |
| Edit-combination EnsemblePLM Backbone=BERT-base-Chinese2023.05 | 42.9 | 20.18 | 35.01 | |
| BARTParameters=400M, Evaluation Protocol=Primary2024.05 | 41.9 | 29.48 | 38.64 | |
| Edit-level EnsemblePLM Backbone=BERT-base-Chinese2023.05 | 41.31 | 21.79 | 35.04 | |
| BaichuanParameters=13B, Evaluation Protocol=LoRA Fine-tuning2024.05 | 37.91 | 26.9 | 35.04 | |
| BaichuanParameters=7B, Evaluation Protocol=LoRA Fine-tuning2024.05 | 36.47 | 25.18 | 33.47 | |
| GPT-4Evaluation Protocol=Zero-shot2024.05 | 36.36 | 27.71 | 34.22 | |
| GPT-4 + DeGLMEvaluation Protocol=Zero-shot with DeGLM detection2024.05 | 32.68 | 30.9 | 32.31 | |
| ChatGLM2Parameters=6B, Evaluation Protocol=LoRA Fine-tuning2024.05 | 31.35 | 21.39 | 28.68 | |
| ChatGLM3Parameters=6B, Evaluation Protocol=LoRA Fine-tuning2024.05 | 30.62 | 21.6 | 28.26 |