ARC
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
ARC-e (leave-one-out setup)
93.5Accuracy (ARC-e LOO)
12
ARC
81.19Verifiability Score
12
ARC
49.45Reusability Score
12
ARC (test)
93Pass@1
12
ARC Hindi
88Accuracy
12
ARC Korean (test)
92Accuracy
12
ARC Easy
86Accuracy
12
ARC Hard
42.41Accuracy
12
ARC (test)
0.848Accuracy
12
ARC Challenge
63.7Accuracy
12
ARC Easy
80.9Accuracy
12
ARC Hard
1.22Accuracy Improvement
12
ARC n=2590 (test)
100Activation
11
ARC n=2390 (test)
100Performance Score (Act)
11
ARC-c zero-shot
40.7Accuracy (zero-shot)
11
ARC-E
87Accuracy (ARC-E)
11
ARC-C
5ECE (15-bin)
11
ARC
86.7Accuracy
11
ARC bo
53.67Accuracy
11
ARC-c
82.51Pass@1
11
ARC 1.2k (test)
1.14MAE
11
ARC-C
28.92Accuracy (Normalized)
11
ARC Bengali (test)
79ARC Accuracy
11
ARC Indonesian 1.0 (test)
91Accuracy
11
ARC Arabic translated (test)
89ARC Score
11