Autonomous Purchasing on staged e-commerce environment
0.983Correct Purchase RateClean Baseline
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Clean BaselineTarget Model=Qwen2.5-VL-7B2026.03 | 0.983 | — | |
| BIMTarget Model=Qwen2.5-VL-7B, epsilon=16/2552026.03 | 0.935 | 0.065 | |
| PGDTarget Model=Qwen2.5-VL-7B, epsilon=16/2552026.03 | 0.923 | 0.077 | |
| Clean BaselineTarget Model=LLaVA-v1.5-7B2026.03 | 0.902 | — | |
| CLIP SpectralTarget Model=Qwen2.5-VL-7B, epsilon=16/2552026.03 | 0.845 | 0.155 | |
| BIMTarget Model=LLaVA-v1.5-7B, epsilon=16/2552026.03 | 0.474 | 0.526 | |
| PGDTarget Model=LLaVA-v1.5-7B, epsilon=16/2552026.03 | 0.462 | 0.538 | |
| CLIP SpectralTarget Model=LLaVA-v1.5-7B, epsilon=16/2552026.03 | 0.331 | 0.669 |