Performance Prediction on WMT En-Fr 2014 (val)
0.28MAELLM-PP
Evaluation results
| Method | Links | ||
|---|---|---|---|
| LLM-PPBackbone=GPT-42023.10 | 0.28 | 0.75 | |
| LLM-PPBackbone=GPT-4, Ablation=+ Fifth instruction2023.10 | 0.28 | 0.75 | |
| LLM-PPBackbone=GPT-4, Ablation=Demonstrations only2023.10 | 0.3 | 0.66 | |
| LLM-PPBackbone=GPT-4, Ablation=+ Second instruction2023.10 | 0.31 | 0.72 | |
| LLM-PPBackbone=GPT-4, Ablation=+ Role + Hyp.2023.10 | 0.32 | 0.71 | |
| LLM-PPBackbone=GPT-4, Ablation=+ Fourth instruction2023.10 | 0.32 | 0.65 | |
| LLM-PPBackbone=GPT-4, Ablation=+ Third instruction2023.10 | 0.33 | 0.73 | |
| LLM-PPBackbone=GPT-4, Ablation=+ First instruction2023.10 | 0.34 | 0.68 | |
| LLM-Distill-PPTeacher Backbone=GPT-42023.10 | 0.34 | 0.76 | |
| LLM-PPBackbone=Mistral2023.10 | 0.6 | 0.34 | |
| LLM-PPBackbone=ChatGPT2023.10 | 0.82 | 0.61 | |
| LLM-Distill-PPTeacher Backbone=ChatGPT2023.10 | 1.01 | 0.79 | |
| Layer-wise MoSType=Baseline2023.10 | 1.16 | 0.79 | |
| Neuron-wise MoSType=Baseline2023.10 | 1.18 | 0.87 | |
| Supernet (Sandwich)Type=Baseline2023.10 | 1.27 | 0.78 | |
| HATType=Baseline2023.10 | 1.59 | 0.79 |