List your company
ResearchFormerly WizWand

Compositional multi-tasking Reply + translation (test)

14.99W-R ScoreJoint-expert LoRA
0.70044.41028.1211.8298Jul 21, 2025
Updated 7mo ago

Evaluation results

MethodLinks
Joint-expert LoRAEfficient?=false2025.0714.9965.73
Learnable Calibration++Efficient?=true2025.0712.1463.81
Multi-step LoRA usageEfficient?=false2025.0710.0469.83
Shared Learnable Calibration++Efficient?=true2025.079.9956.74
Learnable CalibrationEfficient?=true2025.078.8557.46
Shared Learnable CalibrationEfficient?=true2025.076.6349.23
DAM/ZipLoRA mergeEfficient?=true2025.074.6231.88
Auxiliary-task LoRAEfficient?=true2025.074.034.73
Concat mergeEfficient?=true2025.073.9713.1
Linear mergeEfficient?=true2025.073.9412.81
In-context learningEfficient?=true2025.073.938.72
Slerp mergeEfficient?=true2025.073.739.04
TIES mergeEfficient?=true2025.073.538.3
LoraHub mergeEfficient?=true2025.073.2612.69
LM-Cocktail mergeEfficient?=true2025.073.247.93
DARE mergeEfficient?=true2025.072.958.57
Zero-shotEfficient?=true2025.072.034.11
Main-task LoRAEfficient?=true2025.071.257.17