Our new X account is live! Follow @wizwand_team for updates
WorkDL logo mark

Commonsense Reasoning on HellaSwag (Robustness Evaluation)

82.94Accuracy (Baseline)

BF16

24.4439.627554.81570.0025Dec 14, 2025Dec 23, 2025Jan 2, 2026Jan 12, 2026Jan 22, 2026Feb 1, 2026Feb 11, 2026
Updated 4d ago

Evaluation Results

MethodLinks
2026.02
82.94--
2026.02
81.45--1.49
2026.02
81.45--1.49
2026.02
81.38--1.56
2026.02
81.38--1.56
2026.02
80.5--
2026.02
79.71--0.79
2026.02
79.48--1.02
2026.02
79.41--1.09
2026.02
79.15--
2026.02
77.84--1.31
2026.02
77.6--1.55
2026.02
77.47--1.68
2026.02
76.56--2.59
2026.02
75.99--
2026.02
75.02--0.97
2026.02
75--0.99
2026.02
74.92--1.07
2026.02
74.91--1.08
2025.12
62.691.33-
2025.12
62.5926-
2025.12
62.060-
2025.12
61.660-
2025.12
56.690-
2025.12
55.760-
2025.12
55.170-
2025.12
53.550.67-
2025.12
49.3922.66-
2025.12
43.088-
2025.12
43.0422-
2026.02
26.69--53.81