List your company

Toxic Output Mitigation on RealToxicityPrompts 1.0 (Toxic)

0.299ToxicityPREADD-S
0.293320.331660.370.40834Jul 6, 2023

Evaluation results

MethodLinks
PREADD-SBase language model=OPT-6.7B, Sampling strategy=Pure sampling, Generation length=32 tokens, Prefix type=Static2023.070.29969.10.253
PREADD-DBase language model=OPT-6.7B, Sampling strategy=Pure sampling, Generation length=32 tokens, Prefix type=Dynamic2023.070.32675.50.25
FUDGEBase language model=OPT-6.7B, Sampling strategy=Top-k sampling (k=100), Generation length=32 tokens2023.070.35526.60.292
GBase language model=OPT-6.7B, Sampling strategy=Pure sampling, Generation length=32 tokens2023.070.37562.90.272
NEGPROMPTBase language model=OPT-6.7B, Sampling strategy=Pure sampling, Generation length=32 tokens2023.070.44166.30.282