Conversational Classification on Vietnamese Conversational Dataset (test)
97Emotion AccuracyGPT-5
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| GPT-5Adaptation=Prompt-based, Type=Externally hosted LLM2026.01 | 97 | 95 | 84 | 92 | 92 | |
| Credit C-GPTBackbone=Qwen2.5 7B, Adaptation=Domain-specific fine-tuning2026.01 | 90 | 89 | 77 | 88 | 86 | |
| Qwen2.5 7B InstructBackbone=Qwen2.5 7B, Adaptation=Prompt-based, Domain Training=None2026.01 | 78 | 70 | 59 | 62 | 67 | |
| BERT-basedType=Fine-tuned encoder pipeline, Components=Utterance-level intent classifiers, sequence labeling, standalone sentiment analysis2026.01 | 74 | 70 | 65 | 72 | 73 |