List your company

Controllable Text Generation on AGNews (test)

16.94Output Perplexity (O-PPL)GPT2(raw)
15.27426.519537.76549.0105Dec 16, 2022

Evaluation results

MethodLinks
GPT2(raw)Method Category=Zero-shot PLM2022.1216.9474.4152.1746.8845.55
PFMethod Category=Lightweight method2022.1220.2732.3568.4459.1732.73
Ctr-PFMethod Category=Lightweight method2022.1220.4133.983.2160.3431.2
GPT2Method Category=Finetuned PLM2022.1222.2223.4683.0854.9339.93
Ground TruthMethod Category=Reference2022.1231.04—94.8967.2423.31
PT(noise)+PLMethod Category=Self-Training with PLM, Backbone PLM=UniLM, Pseudo Labeler=BERT-base2022.1232.3616.6489.753.7947.95
PT(select)+PLMethod Category=Self-Training with PLM, Backbone PLM=UniLM, Pseudo Labeler=BERT-base2022.1233.2316.6690.5253.7147.69
DuNSTMethod Category=Our method2022.1234.7333.5893.5959.4237.02
T5Method Category=Finetuned PLM2022.1248.3332.1288.4358.0637.01
UniLMMethod Category=Finetuned PLM2022.1255.7936.2887.754.7643.77
PTMethod Category=Self-Training with PLM, Backbone PLM=UniLM2022.1257.440.9586.3652.3546.41
PT(noise)Method Category=Self-Training with PLM, Backbone PLM=UniLM2022.1258.5945.3285.2753.3546.57