Visual Web Navigation and Interaction on VisualWebArena
0.54Success RateSGV
Evaluation Results
| Method | Links | |
|---|---|---|
| SGVHarness=Custom‡, Params=-2026.04 | 0.54 | |
| A3-Qwen3.5-9BHarness=GenericAgent, Params=9B2026.04 | 0.337 | |
| Qwen3.5-9B (base)Harness=GenericAgent, Params=9B2026.04 | 0.262 | |
| Llama-3-70B + SearchHarness=Custom, Params=70B2026.04 | 0.167 | |
| ViGoRLHarness=Custom, Params=7B2026.04 | 0.112 |