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

Downstream Task Evaluation on LAMBADA, HellaSwag, PIQA, WinoGrande, OpenBookQA, and ARC (test)

52.21HellaSwag Accuracy

HiLS-Attn HoPE

23.911631.258338.60545.9517Jul 3, 2026
Updated 18d ago

Evaluation Results

MethodLinks
2026.07
52.2157.0572.1455.6422.570.0237.9745.8628.1449.06
2026.07
51.9957.1370.5156.5122.668.2536.9544.4429.4948.65
2026.07
25-5050252525252525