VLM-as-Judge Evaluation on ArcBench VLM-as-Judge Closed
58.5TQHTML
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| HTMLGeneration Model=GPT-52026.04 | 58.5 | 61.5 | 63.67 | 10 | |
| ArcDeckGeneration Model=Qwen-3-VL-32B-Instruct2026.04 | 53.62 | 63 | 87.25 | 79.5 | |
| HTMLGeneration Model=Qwen-3-VL-32B-Instruct2026.04 | 48.62 | 51.88 | 64 | 24.38 | |
| ArcDeckGeneration Model=GPT-52026.04 | 46.5 | 63.83 | 85.5 | 83.67 | |
| SlideGenGeneration Model=Qwen-3-VL-32B-Instruct2026.04 | 42.5 | 55.25 | 85.25 | 79.5 | |
| P2PosterGeneration Model=Qwen-3-VL-32B-Instruct2026.04 | 36.5 | 49.5 | 72.25 | 75.88 | |
| SlideGenGeneration Model=GPT-52026.04 | 36.17 | 56.67 | 88.33 | 84.83 | |
| P2PosterGeneration Model=GPT-52026.04 | 35.17 | 41.67 | 72 | 78.5 | |
| ArcDeckGeneration Model=GPT-4o2026.04 | 24.38 | 43.25 | 91 | 81.25 | |
| PPTAgentGeneration Model=Qwen-3-VL-32B-Instruct2026.04 | 19.25 | 35.25 | 47.25 | 11.25 | |
| PPTAgentGeneration Model=GPT-52026.04 | 16.5 | 37.67 | 71.33 | 57.5 | |
| PPTAgentGeneration Model=GPT-4o2026.04 | 12.62 | 26.75 | 59.12 | 34.75 | |
| P2PosterGeneration Model=GPT-4o2026.04 | 11.12 | 28.12 | 60.25 | 62.38 | |
| SlideGenGeneration Model=GPT-4o2026.04 | 9.62 | 30.75 | 84.75 | 78.25 | |
| HTMLGeneration Model=GPT-4o2026.04 | 7.25 | 14 | 24 | 0.25 |