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

Super-NaturalInstructions

Benchmarks

Task NameDataset NameSOTA ResultTrend
Instruction FollowingSuper-NaturalInstructions v2 (held-out categories)
Textual Entailment ROUGE-L F164.1
15
Linguistic SteganographySuper-NaturalInstructions
ΔPPL0.669
8
Zero-shot cross-task generalizationSuper-NaturalInstructions (test)
ROUGE-L34.97
8
Cross-task generalizationSuper-NaturalInstructions one-shot
ROUGE-L39.59
4
Showing 4 of 4 rows