VLM-as-Judge Evaluation on ArcBench VLM-as-Judge Open
86.11TQHTML
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| HTMLGeneration Model=GPT-52026.04 | 86.11 | 89.06 | 95.22 | 76.77 | |
| ArcDeckGeneration Model=Qwen-3-VL-32B-Instruct2026.04 | 80.06 | 89 | 98.31 | 94.68 | |
| ArcDeckGeneration Model=GPT-52026.04 | 78.38 | 91.39 | 99.4 | 96.06 | |
| SlideGenGeneration Model=Qwen-3-VL-32B-Instruct2026.04 | 69.18 | 86.4 | 94.16 | 92.18 | |
| SlideGenGeneration Model=GPT-52026.04 | 69.02 | 88.18 | 97.36 | 95.28 | |
| HTMLGeneration Model=Qwen-3-VL-32B-Instruct2026.04 | 67.55 | 71.02 | 82.05 | 66.74 | |
| P2PosterGeneration Model=Qwen-3-VL-32B-Instruct2026.04 | 65.96 | 74.24 | 95.9 | 89.6 | |
| P2PosterGeneration Model=GPT-52026.04 | 62.2 | 70.55 | 97.08 | 91.35 | |
| ArcDeckGeneration Model=GPT-4o2026.04 | 51.68 | 62.78 | 96.05 | 92.94 | |
| PPTAgentGeneration Model=Qwen-3-VL-32B-Instruct2026.04 | 41.55 | 66.12 | 80.44 | 45.25 | |
| PPTAgentGeneration Model=GPT-52026.04 | 35.95 | 55.94 | 84.86 | 81.95 | |
| P2PosterGeneration Model=GPT-4o2026.04 | 32.32 | 40.71 | 92.05 | 86.23 | |
| SlideGenGeneration Model=GPT-4o2026.04 | 19.28 | 41.4 | 93.3 | 89.4 | |
| HTMLGeneration Model=GPT-4o2026.04 | 16.72 | 19.51 | 54.47 | 12.89 | |
| PPTAgentGeneration Model=GPT-4o2026.04 | 16.43 | 25.75 | 82.59 | 71.49 |