Table-based Reasoning and Grounding on HighlightBench Synthetic
53.4T1: Markup Grounding AccuracyOurs
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| OursSubset=Synthetic2026.03 | 53.4 | 88.9 | 73.3 | 81.5 | 78.9 | 72.3 | |
| Gemini2.5-FlashSubset=Synthetic2026.03 | 51.9 | 90.3 | 89.1 | 67.8 | 93.9 | 73.9 | |
| Qwen3-VL-8BSubset=Synthetic2026.03 | 37 | 74.3 | 18.8 | 31.5 | 74.4 | 44.2 | |
| Claude Sonnet 4Subset=Synthetic2026.03 | 32.7 | 81.9 | 40 | 37.5 | 67.1 | 47.6 | |
| InternVL3.5-8BSubset=Synthetic2026.03 | 29 | 81.2 | 17.6 | 24.7 | 71 | 40.5 | |
| MiniCPM-V 4.5Subset=Synthetic2026.03 | 26.2 | 74.3 | 13.3 | 28.8 | 59.2 | 36.9 | |
| MiniCPM-V 2.6Subset=Synthetic2026.03 | 17.6 | 59 | 7.9 | 13.1 | 14.5 | 20.4 | |
| Qwen2.5-VL-3BSubset=Synthetic2026.03 | 14.6 | 61.1 | 6.7 | 13.1 | 47.1 | 24.9 | |
| LLaVA1.5-7BSubset=Synthetic2026.03 | 8.1 | 0.7 | 0 | 0 | 1 | 2.7 |