Conversational Response Generation on Bitext Retail Banking LLM Chatbot (test)
26.85BLEUSFT Model
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| SFT ModelPrompting=0-shot, Alignment=SFT2026.02 | 26.85 | 58.34 | 32.81 | 41.28 | 52.1731 | 91.46 | |
| SFT ModelPrompting=2-shot, Alignment=SFT2026.02 | 25.49 | 57.03 | 30.88 | 39.01 | 53.627 | 91.1 | |
| Base ModelPrompting=2-shot, Alignment=None2026.02 | 2.73 | 34.4 | 9.8 | 17.84 | 37.0881 | 83.39 | |
| Base ModelPrompting=0-shot, Alignment=None2026.02 | 2.7 | 34.18 | 9.08 | 17.87 | 35.1978 | 83.93 | |
| GPT-5Prompting=2-shot2026.02 | 1.37 | 36.33 | 8.06 | 18.69 | 33.6562 | 84.76 |