HellaSwag
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
HellaSwag full (test val)
26.7Accuracy
3
HellaSwag TinyLlama-1.1B-Chat v1.0
50.87HellaSwag Accuracy
3
HellaSwag
0.003RMSE
3
Hellaswag
55.5Performance
3
HellaSwag Spanish (test)
53.1Accuracy
3
HellaSwag first 1000 examples (standard)
38Accuracy (HellaSwag 1k)
3
HellaSwag
93.7Accuracy
3
HellaSwag
82.6Normalized Log Accuracy
3
HellaSwag DE 10-shot (test)
68.7Normalized Log Accuracy
3
HellaSwag Hungarian
51.66Accuracy
3
HellaSwag Turkish
49.6Accuracy
3
HellaSwag
35Error Rate
3
HellaSwag
3.81Average Relative Improvement
3
HellaSwag Multilingual (test)
77.4French Accuracy
3
HellaSwag
0.003MSE
3
HellaSwag
41.1Accuracy
2
Hellaswag fewshot=10 (test)
88.57Accuracy
2
HellaSwag 200 samples
0.67Log-likelihood Score
2
HellaSwag
0.987BPC
2
HellaSwag DE
59.6Normalized Log Accuracy
2
HellaSwag Finance
77Accuracy
2
Hellaswag
4.71Time (h)
2
HellaSwag 77 (test)
—Score
0
HellaSwag Māori
—Accuracy
0