ResearchBenchmarksBenchmark-topology generalisation on WebArena topology simulation seed=42 (n=50 per topology)Benchmark-topology generalisation on WebArena topology simulation seed=42 (n=50 per topology)Follow1.24Rho(B)-reductionWM-SAR0.43920.64710.8551.0629Jul 2, 2026Evaluation resultsMethodMethodLinksRho(B)-reductionMean Region SizeWM-SAR2026.071.2414LR-3Hhops=3hops=32026.071.1411Cascade2026.071.0813Window-4window_size=4window_size=42026.070.894TopK-5K=5K=52026.070.845Greedy-Pt2026.070.591LR-2Hhops=2hops=22026.070.478