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

Robustness Evaluation on R-Bench 2024 (test)

73.53MCQ Accuracy (low)

Robust-U1

31.919642.722353.52564.3277Jun 6, 2026
Updated 1mo ago

Evaluation Results

MethodLinks
2026.06
73.5373.2967.6870.6771.6469.3482.7280.5976.473.98
2026.06
71.7665.8457.9364.9761.2761.546.8546.3342.8857.7
2026.06
65.2963.9160.9749.1449.0949.840.6837.8134.8450.17
2026.06
64.1160.2257.3248.7248.5449.0437.7837.0433.348.45
2026.06
62.3560.2459.1449.8245.3951.0836.6730.4128.5147.06
2026.06
58.2357.7650.648.6546.344.1940.4837.4634.846.49
2026.06
47.0546.5840.2445.0343.3947.4322.922.1919.8337.18
2026.06
46.4742.2340.2446.8739.9444.6121.1121.9519.3735.86
2026.06
33.5226.0830.4826.0722.1224.430.680.650.6718.3