Cross-lingual Aspect-Based Sentiment Analysis on Multilingual ABSA (M-ABSA) (test)
63.23FR ScoreMTL-MSMO-DISTLL
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| MTL-MSMO-DISTLLBackbone=XLM-R2025.02 | 63.23 | 70.95 | 66.24 | 64.36 | 66.2 | |
| Qwen2.5-7BTraining Protocol=LoRA fine-tuning, Fine-tuning Method=LoRA2025.02 | 63.01 | 68.95 | 60.84 | 53.5 | 61.58 | |
| Mistral-7B-v0.3Training Protocol=LoRA fine-tuning, Fine-tuning Method=LoRA2025.02 | 59.46 | 60.79 | 56.89 | 50.52 | 56.92 | |
| MTL-MSMO-DISTLLBackbone=mBERT2025.02 | 54.56 | 62.69 | 55.56 | 56.19 | 57 | |
| Llama-3.1-8BTraining Protocol=LoRA fine-tuning, Fine-tuning Method=LoRA2025.02 | 52.65 | 55.37 | 50.37 | 48.12 | 51.63 | |
| Gemma-2-9b-ItTraining Protocol=Zero-shot prompting, Prompting Template=Wu et al. (2025b)2025.02 | 50.94 | 48.8 | 50.24 | 39.34 | 47.33 | |
| Qwen2.5-7B-InstructTraining Protocol=Zero-shot prompting, Prompting Template=Wu et al. (2025b)2025.02 | 48.88 | 48.29 | 46.75 | 40.25 | 46.04 | |
| GPT-4oTraining Protocol=Zero-shot prompting, Prompting Template=Wu et al. (2025b)2025.02 | 48.43 | 49.91 | 49.94 | 45.15 | 48.36 | |
| Gemma-2-9BTraining Protocol=LoRA fine-tuning, Fine-tuning Method=LoRA2025.02 | 48.17 | 57.46 | 51.97 | 47.65 | 51.31 | |
| Mistral-7B-Instruct-v0.3Training Protocol=Zero-shot prompting, Prompting Template=Wu et al. (2025b)2025.02 | 37.21 | 38.32 | 33.98 | 26.58 | 34.02 | |
| Llama-3.1-8B-InstructTraining Protocol=Zero-shot prompting, Prompting Template=Wu et al. (2025b)2025.02 | 23.15 | 32.49 | 33.53 | 30.18 | 29.84 |