Slide Automation on DynaSlide Closed-Domain
90.12Theme 1 Success RateSlideAgent
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| SlideAgentLLM Backbone=GPT-OSS-120B2026.04 | 90.12 | 71.83 | 78.48 | 77.03 | 81.01 | 85.36 | 80.64 | |
| SlideAgentLLM Backbone=Qwen3-80B2026.04 | 87.65 | 70.14 | 73.42 | 67.53 | 69.62 | 83.64 | 75.33 | |
| SlideAgentLLM Backbone=Qwen3-30B2026.04 | 85.06 | 63.38 | 68.35 | 63.64 | 68.35 | 79.64 | 71.4 | |
| SlideAgentLLM Backbone=GPT-OSS-20B2026.04 | 81.48 | 61.55 | 65.95 | 61.69 | 67.09 | 77.45 | 69.2 | |
| SlideAgentLLM Backbone=Qwen3-14B2026.04 | 50.74 | 50.7 | 34.18 | 38.96 | 45.57 | 52.73 | 45.48 |