HTML observation reduction on WorkArena
0.01Average Wall-Clock Time (seconds)Original (HTML)
Evaluation Results
| Method | Links | |
|---|---|---|
| Original (HTML)Category=Baseline2026.05 | 0.01 | |
| RandomCategory=Baseline2026.05 | 0.1 | |
| GEPACategory=Program-based, ratio=0.62026.05 | 0.1 | |
| GEPACategory=Program-based, ratio=0.22026.05 | 0.1 | |
| DMRCategory=Retrieval-based, variant=BM252026.05 | 0.12 | |
| Pruned by AXTreeCategory=Program-based2026.05 | 0.27 | |
| DMRCategory=Retrieval-based, variant=finetuned2026.05 | 2.55 | |
| DMRCategory=Retrieval-based, variant=Dense2026.05 | 2.95 | |
| DMRCategory=LLM inference, variant=QueryGen2026.05 | 13.79 | |
| Prune4WebCategory=LLM inference2026.05 | 28.32 | |
| FocusAgentCategory=LLM inference2026.05 | 105.69 |