Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Deep Research Evaluation on Deep Research Bench (step 1100)

53.81Readability

Tournament-GRPO

32.261237.855643.4549.0444May 26, 2026
Updated 7d ago

Evaluation Results

MethodLinks
2026.05
53.8155.2255.1955.455.09
2026.05
53.6954.9255.6355.1655.04
2026.05
53.0353.5453.8353.853.64
2026.05
52.9353.953.9654.0453.84
2026.05
52.7953.2652.9553.3853.14
2026.05
52.3853.0453.1752.9152.96
2026.05
51.1551.3450.3850.0550.57
2026.05
51.1151.0150.5250.5250.71
2026.05
50.0650.7650.1649.9450.2
2026.05
49.6550.9250.5350.0150.32
2026.05
48.6949.4347.9446.2447.81
2026.05
45.684646.2346.0146.04
2026.05
41.2641.5641.3340.7741.2
2026.05
40.8838.9236.4235.4737.38
2026.05
40.4938.9836.436.0737.51
2026.05
33.0932.9732.1431.1732.14