List your company
ResearchFormerly WizWand

Performance Prediction on WMT En-De 2019 (val)

0.29MAELLM-PP
0.26320.44410.6250.8059Oct 25, 2023
Updated 8mo ago

Evaluation results

MethodLinks
LLM-PPBackbone=GPT-4, Ablation=+ Third instruction2023.100.290.67
LLM-PPBackbone=GPT-42023.100.320.65
LLM-PPBackbone=GPT-4, Ablation=+ Role + Hyp.2023.100.320.67
LLM-PPBackbone=GPT-4, Ablation=+ Fifth instruction2023.100.320.65
LLM-PPBackbone=GPT-4, Ablation=+ Fourth instruction2023.100.330.71
LLM-PPBackbone=GPT-4, Ablation=Demonstrations only2023.100.340.61
LLM-PPBackbone=GPT-4, Ablation=+ First instruction2023.100.340.58
LLM-PPBackbone=GPT-4, Ablation=+ Second instruction2023.100.350.66
LLM-Distill-PPTeacher Backbone=GPT-42023.100.380.68
LLM-PPBackbone=ChatGPT2023.100.720.56
Neuron-wise MoSType=Baseline2023.100.870.67
HATType=Baseline2023.100.910.72
Supernet (Sandwich)Type=Baseline2023.100.910.72
LLM-PPBackbone=Mistral2023.100.920.18
LLM-Distill-PPTeacher Backbone=ChatGPT2023.100.950.65
Layer-wise MoSType=Baseline2023.100.960.74