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

Multi-task Language Understanding on MMLU-Pro (maj@2, maj@4, maj@8)

54.88Accuracy (maj@2)

DLE (ε-sampling)-RANDBRANCH

Updated 1mo ago

Evaluation Results

Method	Links
DLE (ε-sampling)-RANDBRANCH 2026.04		54.88	56.52	58.29
DLE (ε-sampling)-DIVFIRST 2026.04		54.66	56.4	58.18
DLE (top-p+top-k) 2026.04		54.5	56.27	58.06
DLE (min-p) 2026.04		54.34	56.13	58.23
DLE (ε-sampling)-PROBFIRST 2026.04		54.33	56.38	58.3
Self-consistency (ε-sampling) 2026.04		52.94	56.48	58.56
Self-consistency (top-p+top-k) 2026.04		52.68	55.31	57.74
Self-consistency (min-p) 2026.04		52.68	55.31	57.74
Self-consistency 2026.04		52.01	55.5	58.39
DLE 2026.04		20.71	22.17	22.85
DLE 2026.04		20.62	22.17	22.86
DLE 2026.04		19.76	21.23	22.15
DLE 2026.04		19.75	20.9	21.97
DLE 2026.04		19.61	21.13	22.95
Self-consistency 2026.04		18.87	20.69	21.82
Self-consistency 2026.04		17.73	20.09	21.29
Self-consistency 2026.04		17.3	19.3	20.78
Self-consistency 2026.04		13.73	16.71	19.02