Slide Editing on TSBench SlideStruct
91.11Success Rate (SR)TALK-TO-YOUR-SLIDES
Evaluation Results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| TALK-TO-YOUR-SLIDESBackbone=GPT-4.1-mini2025.05 | 91.11 | 1.4 | 2.44 | 2.59 | 2.78 | 2.9 | 74.39 | 2.24 | 1.28 | 2.1 | |
| TALK-TO-YOUR-SLIDESEvaluation Model=Claude 3.5 Haiku2025.05 | 90.42 | 1.37 | 2.39 | 2.54 | 2.73 | 2.84 | 72.1 | 2.19 | 1.25 | 1.9 | |
| Direct code gen.Backbone=GPT-4.1-mini2025.05 | 84.44 | 0.17 | 1.78 | 1.11 | 1.91 | 1.64 | 19.57 | 0.91 | 0.65 | 1.4 | |
| Direct code gen.Evaluation Model=Claude 3.5 Haiku2025.05 | 83.72 | 0.16 | 1.73 | 1.09 | 1.86 | 1.6 | 18.94 | 0.9 | 0.63 | 1.2 | |
| UI AgentBackbone=GPT-4.1-mini2025.05 | 82.3 | 1.55 | 2.27 | 1.3 | 2.08 | 1.99 | 95.07 | 75.45 | 1.77 | 11.9 | |
| UI AgentEvaluation Model=Claude 3.5 Haiku2025.05 | 81.55 | 1.5 | 2.2 | 1.27 | 2.03 | 1.93 | 92.01 | 71.9 | 1.7 | 9.9 |