Table Reasoning on MMTU (test)
56AccuracyCPO
Evaluation Results
| Method | Links | |
|---|---|---|
| CPOModel=Qwen3-14B-Base2026.06 | 56 | |
| Table-R1Model=Table-R12026.06 | 51 | |
| DPOModel=Mistral-8B-Instruct2026.06 | 49.8 | |
| CPOModel=Mistral-8B-Instruct2026.06 | 46 | |
| Trace SFTModel=Qwen3-14B-Base2026.06 | 44 | |
| CPOModel=Llama-3.1-8B-Instruct2026.06 | 43 | |
| Trace SFTModel=Llama-3.1-8B-Instruct2026.06 | 40 | |
| DPOModel=Llama-3.1-8B-Instruct2026.06 | 38.3 | |
| Trace SFTModel=Mistral-8B-Instruct2026.06 | 33 | |
| Q&A SFTModel=Llama-3.1-8B-Instruct2026.06 | 31 | |
| Q&A SFTModel=Mistral-8B-Instruct2026.06 | 30 | |
| Q&A SFTModel=Qwen3-14B-Base2026.06 | 29 | |
| DPOModel=Qwen3-14B-Base2026.06 | 24.2 |