Scenario classification on Customer-service conversations
71.39Scenario F1ELECTRA
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| ELECTRAMethod=EARS2026.06 | 71.39 | 74.96 | 88.61 | |
| RoBERTaMethod=EARS2026.06 | 70.33 | 73.72 | 89.81 | |
| DeBERTaMethod=EARS2026.06 | 69.83 | 75.17 | 91.04 | |
| RoBERTaMethod=Transformer2026.06 | 66.15 | 71.39 | 88.88 | |
| DeBERTaMethod=Transformer2026.06 | 65.39 | 71.21 | 91.64 | |
| ELECTRAMethod=Transformer2026.06 | 64.89 | 70.13 | 87.71 | |
| GPT-5.4Method=LLM Few-shot2026.06 | 45.77 | 54.23 | 81.9 | |
| GPT-5.4Method=LLM Zero-shot2026.06 | 42.52 | 52.61 | 78.35 | |
| Qwen3-32BMethod=LLM Few-shot2026.06 | 42.14 | 50.84 | 73.07 | |
| GPT-OSS-20BMethod=LLM Few-shot2026.06 | 34.59 | 47.29 | 68.62 | |
| Qwen3-32BMethod=LLM Zero-shot2026.06 | 33.15 | 45.88 | 68.12 | |
| Llama-3.1-8BMethod=LLM Few-shot2026.06 | 30.74 | 41.52 | 65.23 | |
| GPT-OSS-20BMethod=LLM Zero-shot2026.06 | 29.44 | 44.11 | 57.48 | |
| Llama-3.1-8BMethod=LLM Zero-shot2026.06 | 15.56 | 32.15 | 47.52 |