Wealth Management Chatbot Response Generation on Bitext Wealth Management LLM Chatbot (test)
0.2692BLEUDPO Model
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| DPO ModelBackbone=Qwen3-4B-Instruct-2507, Alignment=DPO, Toolkit=AlignTune2026.02 | 0.2692 | 0.5795 | 0.3281 | 0.4113 | 51.5908 | 0.9142 | |
| SFT ModelBackbone=Qwen3-4B-Instruct-2507, Alignment=SFT, Toolkit=AlignTune2026.02 | 0.269 | 0.5796 | 0.327 | 0.4182 | 51.8492 | 0.9134 | |
| GPT-5 (2-shot)Shots=22026.02 | 0.1218 | 0.4703 | 0.1805 | 0.2704 | 46.0776 | 0.8762 | |
| GPT-4o (2-shot)Shots=22026.02 | 0.0705 | 0.4062 | 0.1307 | 0.2323 | 37.4132 | 0.8676 | |
| Base Model (2-shot)Backbone=Qwen3-4B-Instruct-2507, Shots=2, Alignment=Base2026.02 | 0.0398 | 0.3726 | 0.1025 | 0.1968 | 37.0617 | 0.8443 | |
| Base Model (0-shot)Backbone=Qwen3-4B-Instruct-2507, Shots=0, Alignment=Base2026.02 | 0.0286 | 0.3312 | 0.0846 | 0.1729 | 34.3816 | 0.8343 |