Table Fact Verification on TFV TabFact InfoTabs
88.16TabFact AccuracyGemini-2.5-Flash
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Gemini-2.5-FlashModel Category=Closed-Source VLMs2026.04 | 88.16 | 91.43 | 89.8 | |
| V-tableR1 (4B)Model Category=Ours, Parameter Scale=4B2026.04 | 87.95 | 88.94 | 88.45 | |
| GPT-4.1Model Category=Closed-Source VLMs2026.04 | 87.32 | 92.43 | 89.88 | |
| GPT-4oModel Category=Closed-Source VLMs2026.04 | 85.16 | 88.75 | 86.96 | |
| V-tableR1 (2B)Model Category=Ours, Parameter Scale=2B2026.04 | 84.69 | 87.54 | 86.12 | |
| Qwen3-VL-4B (SFT)Model Category=Ours, Parameter Scale=4B, Fine-tuning Protocol=SFT2026.04 | 83.88 | 87.33 | 85.61 | |
| Qwen3-VL-32BModel Category=Open-Source VLMs, Parameter Scale=32B2026.04 | 82.43 | 78.25 | 80.34 | |
| Qwen3-VL-8BModel Category=Open-Source VLMs, Parameter Scale=8B2026.04 | 79.44 | 71.33 | 75.39 | |
| Qwen3-VL-2B (SFT)Model Category=Ours, Parameter Scale=2B, Fine-tuning Protocol=SFT2026.04 | 79.28 | 85.17 | 82.23 | |
| QVQ-72BModel Category=Open-Source VLMs, Parameter Scale=72B2026.04 | 77.68 | 74.64 | 76.16 | |
| InternVL-2.5-14BModel Category=Open-Source VLMs, Parameter Scale=14B2026.04 | 75.98 | 77.48 | 76.73 | |
| Qwen2.5-VL-72BModel Category=Open-Source VLMs, Parameter Scale=72B2026.04 | 73.45 | 72.82 | 73.14 | |
| Qwen3-VL-4BModel Category=Open-Source VLMs, Parameter Scale=4B2026.04 | 73.28 | 77.33 | 75.31 | |
| InternVL-2.5-8BModel Category=Open-Source VLMs, Parameter Scale=8B2026.04 | 72.92 | 74.67 | 73.8 | |
| Qwen3-VL-2BModel Category=Open-Source VLMs, Parameter Scale=2B2026.04 | 70.4 | 74.83 | 72.62 | |
| Table-LLaVA-13BModel Category=Open-Source VLMs, Parameter Scale=13B2026.04 | 65 | 66.91 | 65.96 | |
| Table-LLaVA-7BModel Category=Open-Source VLMs, Parameter Scale=7B2026.04 | 59.85 | 65.26 | 62.56 | |
| LLaVA-NeXT-7BModel Category=Open-Source VLMs, Parameter Scale=7B2026.04 | 34.18 | 41.23 | 37.71 | |
| LLaVA-1.5-13BModel Category=Open-Source VLMs, Parameter Scale=13B2026.04 | 28.37 | 37.34 | 32.86 | |
| LLaVA-1.5-7BModel Category=Open-Source VLMs, Parameter Scale=7B2026.04 | 23.25 | 32.34 | 27.8 |