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

Self-Instruct

Benchmarks

Task NameDataset NameSOTA ResultTrend
Instruction FollowingSelf-Instruct
ROUGE-L16.5
48
Instruction FollowingSelf-Instruct (test)
ROUGE-L21.81
42
Instruction FollowingSelf-instruct Eval
Win Rate (A)56.7
19
Language GenerationSelf-Instruct (test)
ROUGE-L23.4
14
Prompt RecoverySelf-Instruct
BLEU-134.71
14
Open-ended instruction followingSelf-instruct Eval
Win Rate (A)53.6
7
Instruction Following EvaluationSELF-INSTRUCT Ours
Score74.29
5
Instruction Following EvaluationSELF-INSTRUCT
Score69.48
5
Instruction Following EvaluationSELF-INSTRUCT seed data
Score72.01
5
Instruction Following EvaluationSelf-instruct Eval
Win Rate (A)56.7
5
Instruction FollowingSelf-Instruct
Language Democratization73.92
4
Showing 11 of 11 rows