Web navigation on Mind2Web-SC
33.3Task Success Rate (TSR)baseline
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| baselineAccess Control=LLM-AC, Core LLM=GPT-4o2026.06 | 33.3 | 3 | |
| FragFuseAccess Control=AGrail, Core LLM=GPT-4o2026.06 | 23.6 | 72 | |
| baselineAccess Control=AGrail, Core LLM=GPT-5.12026.06 | 22.2 | 9 | |
| direct queryingAccess Control=None, Core LLM=GPT-4o2026.06 | 22 | — | |
| FragFuseAccess Control=AGrail, Core LLM=GPT-5.12026.06 | 20.9 | 86 | |
| FragFuseAccess Control=LLM-AC, Core LLM=GPT-4o2026.06 | 20.5 | 88 | |
| FragFuseAccess Control=LLM-AC, Core LLM=GPT-5.12026.06 | 20.4 | 98 | |
| FragFuseAccess Control=AGrail, Core LLM=Average2026.06 | 20.1 | 82.3 | |
| baselineAccess Control=AGrail, Core LLM=GPT-4o2026.06 | 20 | 5 | |
| FragFuseAccess Control=LLM-AC, Core LLM=Average2026.06 | 19.4 | 93 | |
| direct queryingAccess Control=None, Core LLM=Average2026.06 | 18 | — | |
| FragFuseAccess Control=LLM-AC, Core LLM=Gemini 2.5 Flash2026.06 | 17.2 | 93 | |
| direct queryingAccess Control=None, Core LLM=GPT-5.12026.06 | 17 | — | |
| baselineAccess Control=AGrail, Core LLM=Average2026.06 | 16.6 | 9 | |
| FragFuseAccess Control=AGrail, Core LLM=Gemini 2.5 Flash2026.06 | 15.7 | 89 | |
| direct queryingAccess Control=None, Core LLM=Gemini 2.5 Flash2026.06 | 15 | — | |
| baselineAccess Control=LLM-AC, Core LLM=Average2026.06 | 11.1 | 3 | |
| baselineAccess Control=AGrail, Core LLM=Gemini 2.5 Flash2026.06 | 7.7 | 13 | |
| baselineAccess Control=LLM-AC, Core LLM=GPT-5.12026.06 | 0 | 1 | |
| baselineAccess Control=LLM-AC, Core LLM=Gemini 2.5 Flash2026.06 | 0 | 5 |