List your company
ResearchFormerly WizWand

Compositional Multi-tasking on Compositional Summation and Translation (test)

28.64R-L Score (%)Learnable Calibration++
6.64412.354518.06523.7755Jul 21, 2025
Updated 7mo ago

Evaluation results

MethodLinks
Learnable Calibration++Efficient?=true2025.0728.6465.15
Learnable CalibrationEfficient?=true2025.0725.459.23
Joint-expert LoRAEfficient?=false2025.0721.3649.85
Multi-step LoRA usageEfficient?=false2025.0721.2572.92
Shared Learnable Calibration++Efficient?=true2025.0719.2332.61
Shared Learnable CalibrationEfficient?=true2025.0717.0429.91
DAM/ZipLoRA mergeEfficient?=true2025.0716.1917.16
In-context learningEfficient?=true2025.0714.4610.95
Concat mergeEfficient?=true2025.0714.390.34
Linear mergeEfficient?=true2025.0714.270.33
Auxiliary-task LoRAEfficient?=true2025.0713.990.3
Slerp mergeEfficient?=true2025.0713.960.54
LoraHub mergeEfficient?=true2025.0713.781.59
LM-Cocktail mergeEfficient?=true2025.0713.620.7
Main-task LoRAEfficient?=true2025.0713.393.49
TIES mergeEfficient?=true2025.0712.250.81
DARE mergeEfficient?=true2025.079.270.68
Zero-shotEfficient?=true2025.077.490.44