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

Open-ended writing on DeepresearchBench

46.93Overall Score

Qwen3-8B + R2-Write-SFT + RLp

37.757240.138642.5244.9014Apr 3, 2026
Updated 13d ago

Evaluation Results

MethodLinks
2026.04
46.9344.1644.0349.9849.54
2026.04
4643.3242.8648.7549.08
2026.04
44.9743.843.0345.6247.43
2026.04
44.9542.2442.0846.6848.79
2026.04
44.6943.6642.445.2147.48
2026.04
44.1842.3141.8947.5647.06
2026.04
42.4241.239.4444.1844.86
2026.04
40.5238.7436.2745.8544.18
2026.04
39.7440.2537.841.2239.68
2026.04
38.6836.8434.4543.1840.26
2026.04
38.1136.3233.4143.6642.33