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

Zero-shot Classification on Evaluation Suite (PiQA, LAMBDA, ARC-e, ARC-c, HellaS)

5.23Decode Latency

Longformer + SFA (k=8)

4.7567.955511.15514.3545Mar 17, 2026
Updated 24d ago

Evaluation Results

MethodLinks
2026.03
5.236.1840.7520.5426.396.6317.2422.31
2026.03
6.757.9341.2821.2728.027.0118.9223.3
2026.03
6.929.4139.9920.2427.136.8317.7422.39
2026.03
7.928.3741.3921.0828.317.1118.8923.35
2026.03
9.099.4140.8320.0427.857.1318.0322.78
2026.03
9.437.9339.8319.1126.726.7715.3821.56
2026.03
9.8816.8642.4921.9228.438.0119.3824.05
2026.03
10.8416.8642.3420.7928.37.8218.8323.62
2026.03
11.3916.8642.121.2928.017.9919.2423.73
2026.03
13.3216.8641.8120.5527.047.3818.7523.11
2026.03
14.129.4141.6221.0328.417.3919.2623.54
2026.03
17.0816.8642.7422.7828.358.1219.6124.32