Resource Hijacking Evaluation on GPQA (test)
83,717Research And Data ScoreLeechHijack
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| LeechHijackAgent=Hybrid (OWL)2025.12 | 83,717 | 117,610 | 97,032 | — | 106,473 | |
| LeechHijackAgent=Cloud-Hosted (OpenManus)2025.12 | 63,739 | 141,109 | 75,753 | — | 75,753 | |
| NormalAgent=Cloud-Hosted (OpenManus)2025.12 | 62,656 | 69,294 | 49,638 | — | 49,638 | |
| NormalAgent=Hybrid (OWL)2025.12 | 51,362 | 95,635 | 80,932 | — | 89,124 | |
| NormalAgent=Localized (OpenManus)2025.12 | 44,870 | 85,699 | 44,401 | — | 59,825 | |
| LeechHijackAgent=Localized (OpenManus)2025.12 | 37,480 | 104,656 | 82,366 | — | 61,186 | |
| LeechHijackAgent=Hybrid (Pydantic-AI)2025.12 | 31,353 | 108,585 | 34,013 | — | 34,013 | |
| NormalAgent=Hybrid (Pydantic-AI)2025.12 | 19,356 | 82,389 | 24,541 | — | 24,541 | |
| Extra TaskAgent=Hybrid (OWL)2025.12 | 13,196 | 4,602 | 20,802 | — | 23,264 | |
| Extra TaskAgent=Hybrid (Pydantic-AI)2025.12 | 7,149 | 4,819 | 7,282 | — | 7,283 | |
| Extra TaskAgent=Cloud-Hosted (OpenManus)2025.12 | 4,668 | 5,067 | 10,501 | — | 14,658 | |
| Extra TaskAgent=Localized (OpenManus)2025.12 | 4,560 | 5,227 | 7,740 | — | 6,229 |