Parallel Execution Prediction on EPIC 2-Body Problem (test)
100CoverageNaive Half-Half
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Naive Half-HalfScheduling Strategy=Half-Half2025.12 | 100 | 100 | 2 | 29.1 | 0 | |
| HEFT 1-minScheduling Strategy=HEFT, Time Granularity=1-min2025.12 | 99.5 | 100 | 1.96 | 44.2 | 0.63 | |
| Gemini 2.5 + Spatial PromptPrompting Strategy=Base Prompt + Spatial Prompt (ours)2025.12 | 89.9 | 90.6 | 1.35 | 10.1 | 0.43 | |
| Gemini 2.5 + Goals-and-ConstraintsPrompting Strategy=Base Prompt + Goals-and-Constraints2025.12 | 80.1 | 80 | 1.57 | 21.3 | 0.37 | |
| Gemini 2.5 + Goals-OnlyPrompting Strategy=Base Prompt + Goals-Only2025.12 | 75.8 | 76.9 | 1.54 | 18.8 | 0.44 | |
| HEFT GT start-endScheduling Strategy=HEFT, Ground-truth Action Segments=true2025.12 | 73.9 | 98.5 | 1.7 | 40.1 | 0.36 | |
| Gemini 2.5 (Base Prompt)Prompting Strategy=Base Prompt2025.12 | 55.2 | 55.3 | 1.52 | 22.3 | 0.52 | |
| Qwen2.5-VL-72BNotes=subset of 7/20 videos2025.12 | 41 | 37.8 | 0.89 | 0.3 | 0.12 |