Action Prediction on Human Evaluation User Actions Dataset (test)
79Win RateLongNAP
Evaluation Results
| Method | Links | |
|---|---|---|
| LongNAPModel=Qwen 2.5 VL2026.03 | 79 | |
| Few-shot RAGModel=Gemini2026.03 | 58.5 | |
| Zero-shotModel=Gemini2026.03 | 55 | |
| Few-shot RAGModel=Qwen 2.5 VL2026.03 | 45 | |
| Zero-shotModel=Qwen 2.5 VL2026.03 | 33 | |
| SFTModel=Qwen 2.5 VL2026.03 | 29.5 |