Multi-table Reasoning on BIRD 3-table
56.25AccuracyCPO
Evaluation Results
| Method | Links | |
|---|---|---|
| CPOModel=Qwen3-14B2026.06 | 56.25 | |
| CPOModel=Llama-3.1-8B2026.06 | 56.25 | |
| Q&A SFTModel=Qwen3-14B2026.06 | 50 | |
| Trace SFTModel=Qwen3-14B2026.06 | 50 | |
| CPOModel=Mistral-8B2026.06 | 43.75 | |
| Trace SFTModel=Llama-3.1-8B2026.06 | 43.75 | |
| Q&A SFTModel=Mistral-8B2026.06 | 31.25 | |
| Q&A SFTModel=Llama-3.1-8B2026.06 | 31.25 | |
| Trace SFTModel=Mistral-8B2026.06 | 18.75 |