VLM Pairwise Preference on ArcBench
100Win RateArcDeck
Evaluation Results
| Method | Links | |
|---|---|---|
| ArcDeckGeneration Model=GPT-5, Evaluation Model=Qwen3, Baseline=PPTAgent2026.04 | 100 | |
| ArcDeckGeneration Model=GPT-5, Evaluation Model=GPT-5, Baseline=PPTAgent2026.04 | 100 | |
| ArcDeckGeneration Model=Qwen3, Evaluation Model=Qwen3, Baseline=PPTAgent2026.04 | 100 | |
| ArcDeckGeneration Model=GPT-4o, Evaluation Model=Qwen3, Baseline=PPTAgent2026.04 | 99 | |
| ArcDeckGeneration Model=Qwen3, Evaluation Model=GPT-5, Baseline=PPTAgent2026.04 | 97.9 | |
| ArcDeckGeneration Model=GPT-5, Evaluation Model=GPT-5, Baseline=P2Poster2026.04 | 97.6 | |
| ArcDeckGeneration Model=GPT-4o, Evaluation Model=GPT-5, Baseline=PPTAgent2026.04 | 95 | |
| ArcDeckGeneration Model=Qwen3, Evaluation Model=Qwen3, Baseline=P2Poster2026.04 | 92.8 | |
| ArcDeckGeneration Model=GPT-4o, Evaluation Model=GPT-5, Baseline=HTML2026.04 | 92 | |
| ArcDeckGeneration Model=Qwen3, Evaluation Model=GPT-5, Baseline=P2Poster2026.04 | 90.7 | |
| ArcDeckGeneration Model=GPT-5, Evaluation Model=GPT-5, Baseline=HTML2026.04 | 84.7 | |
| ArcDeckGeneration Model=Qwen3, Evaluation Model=GPT-5, Baseline=HTML2026.04 | 84.5 | |
| ArcDeckGeneration Model=GPT-4o, Evaluation Model=GPT-5, Baseline=P2Poster2026.04 | 84 | |
| ArcDeckGeneration Model=GPT-5, Evaluation Model=Qwen3, Baseline=HTML2026.04 | 82.4 | |
| ArcDeckGeneration Model=GPT-4o, Evaluation Model=Qwen3, Baseline=SlideGen2026.04 | 81 | |
| ArcDeckGeneration Model=GPT-4o, Evaluation Model=Qwen3, Baseline=HTML2026.04 | 79 | |
| ArcDeckGeneration Model=GPT-4o, Evaluation Model=Qwen3, Baseline=P2Poster2026.04 | 78 | |
| ArcDeckGeneration Model=Qwen3, Evaluation Model=Qwen3, Baseline=HTML2026.04 | 76.3 | |
| ArcDeckGeneration Model=GPT-4o, Evaluation Model=GPT-5, Baseline=SlideGen2026.04 | 68 | |
| ArcDeckGeneration Model=GPT-5, Evaluation Model=Qwen3, Baseline=P2Poster2026.04 | 62.4 | |
| ArcDeckGeneration Model=Qwen3, Evaluation Model=Qwen3, Baseline=SlideGen2026.04 | 57.7 | |
| ArcDeckGeneration Model=GPT-5, Evaluation Model=Qwen3, Baseline=SlideGen2026.04 | 55.3 | |
| ArcDeckGeneration Model=GPT-5, Evaluation Model=GPT-5, Baseline=SlideGen2026.04 | 50.6 | |
| ArcDeckGeneration Model=Qwen3, Evaluation Model=GPT-5, Baseline=SlideGen2026.04 | 48.5 | |
| ArcDeckGeneration Model=GPT-5, Evaluation Model=GPT-52026.04 | 48.1 | |
| ArcDeckGeneration Model=Qwen3, Evaluation Model=Qwen32026.04 | 47.3 | |
| SlideGenGeneration Model=GPT-5, Evaluation Model=Qwen32026.04 | 46.8 | |
| ArcDeckGeneration Model=Qwen3, Evaluation Model=GPT-52026.04 | 46.7 | |
| SlideGenGeneration Model=Qwen3, Evaluation Model=Qwen32026.04 | 46 | |
| SlideGenGeneration Model=GPT-5, Evaluation Model=GPT-52026.04 | 45.8 | |
| SlideGenGeneration Model=Qwen3, Evaluation Model=GPT-52026.04 | 45.8 | |
| PPTAgentGeneration Model=GPT-5, Evaluation Model=Qwen32026.04 | 45.6 | |
| PPTAgentGeneration Model=Qwen3, Evaluation Model=Qwen32026.04 | 45.4 | |
| HTMLGeneration Model=Qwen3, Evaluation Model=Qwen32026.04 | 42.6 | |
| P2PosterGeneration Model=Qwen3, Evaluation Model=GPT-52026.04 | 40 | |
| PPTAgentGeneration Model=GPT-5, Evaluation Model=GPT-52026.04 | 40 | |
| HTMLGeneration Model=Qwen3, Evaluation Model=GPT-52026.04 | 40 | |
| ArcDeckGeneration Model=GPT-4o, Evaluation Model=GPT-52026.04 | 40 | |
| P2PosterGeneration Model=Qwen3, Evaluation Model=Qwen32026.04 | 37.7 | |
| PPTAgentGeneration Model=Qwen3, Evaluation Model=GPT-52026.04 | 36.7 | |
| ArcDeckGeneration Model=GPT-4o, Evaluation Model=Qwen32026.04 | 36.6 | |
| HTMLGeneration Model=GPT-5, Evaluation Model=Qwen32026.04 | 35.9 | |
| ArcDeckGeneration Model=GPT-5, Evaluation Model=Qwen32026.04 | 35.7 | |
| PPTAgentGeneration Model=GPT-4o, Evaluation Model=Qwen32026.04 | 35.4 | |
| P2PosterGeneration Model=GPT-5, Evaluation Model=Qwen32026.04 | 35.1 | |
| P2PosterGeneration Model=GPT-5, Evaluation Model=GPT-52026.04 | 33.3 | |
| SlideGenGeneration Model=GPT-4o, Evaluation Model=GPT-52026.04 | 33.3 | |
| HTMLGeneration Model=GPT-5, Evaluation Model=GPT-52026.04 | 33.3 | |
| HTMLGeneration Model=GPT-4o, Evaluation Model=GPT-52026.04 | 33.3 | |
| SlideGenGeneration Model=GPT-4o, Evaluation Model=Qwen32026.04 | 32.1 | |
| HTMLGeneration Model=GPT-4o, Evaluation Model=Qwen32026.04 | 31.3 | |
| PPTAgentGeneration Model=GPT-4o, Evaluation Model=GPT-52026.04 | 30 | |
| P2PosterGeneration Model=GPT-4o, Evaluation Model=Qwen32026.04 | 27.5 | |
| P2PosterGeneration Model=GPT-4o, Evaluation Model=GPT-52026.04 | 26.7 |