Video-based Multi-Step Operational Knowledge Extraction on Mobile Screen Teach Benchmark
56.96PrecisionTeach VLM
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Teach VLM2026.06 | 56.96 | 53.68 | 55.27 | 43.03 | |
| Doubao-Seed-2.02026.06 | 45.25 | 49.28 | 47.18 | 54.75 | |
| Qwen3-VL-8B2026.06 | 37.53 | 37.15 | 37.34 | 62.47 | |
| Doubao-Seed-1.82026.06 | 37.42 | 41.38 | 39.3 | 62.58 | |
| GPT-4o2026.06 | 30.45 | 33.7 | 32 | 69.55 | |
| Qwen3.5-9B2026.06 | 1.7 | 20.91 | 3.15 | 98.3 |