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

Instruction Following on Inverse IFEval, MARS-Bench, MultiChallenge, Collie-Hard, EIFBench

99Pass@1

GPT-5 High

42.52857.18971.8586.511Mar 21, 2026
Updated 25d ago

Evaluation Results

MethodLinks
2026.03
99
2026.03
95.5
2026.03
80.8
2026.03
80.6
2026.03
80.3
2026.03
78.9
2026.03
77.6
2026.03
77.2
2026.03
75.3
2026.03
73.6
2026.03
72.6
2026.03
72.5
2026.03
70.2
2026.03
70.1
2026.03
69.6
2026.03
69.5
2026.03
67.4
2026.03
66.7
2026.03
66.7
2026.03
57.2
2026.03
55.4
2026.03
50.1
2026.03
48.6
2026.03
47
2026.03
44.7