Table Question Answering on TAT-QA (test)
60.75AccuracyMiniCPM-V-2.6 8B
Evaluation results
| Method | Links | |
|---|---|---|
| MiniCPM-V-2.6 8BReasoning Strategy=HIPPO, # TIR=55.2K2026.02 | 60.75 | |
| GPT-42025.01 | 55.81 | |
| TAMAbackbone=LLaMA 3.1 8B Instruct2025.01 | 48.47 | |
| Qwen2-VL-7B-Table-R1Reasoning Strategy=GRPO, # TIR=20.6K2026.02 | 48.06 | |
| Qwen3-VL-8B-DISCOReasoning Strategy=Table-GLS, # TIR=02026.02 | 40.54 | |
| Qwen3-VL-8B-DISCOReasoning Strategy=DA, # TIR=02026.02 | 37.95 | |
| Qwen3-VL-8BReasoning Strategy=DA, # TIR=02026.02 | 36.01 | |
| Gemma3n-E4B-DISCOReasoning Strategy=Table-GLS, # TIR=02026.02 | 33.16 | |
| GPT-3.52025.01 | 31.38 | |
| GPT-4o-miniReasoning Strategy=DA, # TIR=02026.02 | 28.37 | |
| LLaMA 3.1 8B Instructstatus=base model2025.01 | 26.7 | |
| Gemma3n-E4B-DISCOReasoning Strategy=DA, # TIR=02026.02 | 25.26 | |
| Gemma3n-E4BReasoning Strategy=DA, # TIR=02026.02 | 22.41 | |
| Table-LLaVA 13BReasoning Strategy=SFT, # TIR=82K2026.02 | 15.67 | |
| Table-LLaVA 7BReasoning Strategy=SFT, # TIR=82K2026.02 | 12.82 |