LITMUS Predictor

What is the pretrained model you are using for your task?
What is the nature of your task?

Upload file with your observations: CSV containing tuples of form:
[, pivotLang],[, trainingSize],targetLang,metricValue

Selected Fine-Tuning Configurations:

Add fine-tuning data

Steps:

Select training language from the drop-down.
Configure the amount of labelled data in it using the slider.
Click "Add" to save the current training language to the "Current Configuration".
Repeat for each training language in current configuration.
Click "Save" to save the current training configuration.

Language for which you have data (pivots)

Task data size for pivot languages: 100

Current Configuration:

No configurations selected.

Target Languages: Comma-separated list (ex: en,hi,ta)
Do you want data collection strategies for your model?

Edit budget parameters

Budget settings

Languages for which you can label additional training examples: Comma-separated list (ex: en,hi,ta)

Total no. of additional instances that you can label for the selected target languages:

Language-specific budgets: Comma-separated list of key:value pairs (Ex en:5000,fr:2000)

Which config do you want to see data-collection strategies for:

Edit objective functions

Objective function settings

What do you want to maximize?

Weighted Average

Minimum

Target weights for weighted average objective: Comma-separated list of key:value pairs (ex: en:2,fr:3)

Edit performance constraints

Performance constraint settings

Minimum acceptable average performance across tgts:

Minimum acceptable performance per target language: Comma-separated list of key:value pairs (Ex en:0.5,fr:0.7)

Predicted Performance in selected targets:
(reported for best configuration = Config-)

*For predicted perf. on remaining targets, download all results from below.

Predicted Performance in selected targets:
(reported for best configuration = Config-)

Target-Performance Heatmap for different configurations:

Data-collection plan:
(reported for configuration = Config-)

Predicted Performance in selected targets: (reported for best configuration = Config-)

Target-Performance Heatmap for different configurations:

Data-collection plan: (reported for configuration = Config-)

Predicted Performance in selected targets:
(reported for best configuration = Config-)

Data-collection plan:
(reported for configuration = Config-)