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

AlpacaFarm

Benchmarks

Task NameDataset NameSOTA ResultTrend
Indirect Prompt InjectionAlpacaFarm (test)
Attack Success Rate0
105
Instruction FollowingAlpacaFarm (test)
Reward Score387.196
40
Direct Prompt InjectionAlpacaFarm (208 samples)
Naive Success Rate78.36
30
Instruction FollowingAlpacaFarm Eval (test)
Win Rate76.13
28
Instruction FollowingAlpacaFarm
Win Rate59.2
27
Generation quality evaluationAlpacaFarm
Win Rate36.4
12
Showing 6 of 6 rows