Unclear in general how to use classification models requiring sentence pairs

[alex2awesome]

Link to the documentation pages (if available)

There are many NLP tasks that classify pairs of text. E.g. NLI, Stance Detection, Question-Similarity.

It's very unclear how to use these models, many of which exist on SparkNLPHub, in a real pipeline.

Here's an example of great documentation: https://sparknlp.org/2021/08/25/classifierdl_electra_questionpair_pipeline_en.html

it's clear how to format q1 and q2.

However, here's an example of bad documentation:
https://sparknlp.org/2023/09/18/albert_xxlarge_v2_snli_mnli_fever_anli_r1_r2_r3_nli_en.html
https://sparknlp.org/2023/11/29/roberta_classifier_large_snli_mnli_fever_anli_r1_r2_r3_nli_en.html

The first has no string mentioned, even. The second just says "PUT YOUR STRING HERE". I have no idea how to format my input.

How could the documentation be improved?

In general, having a uniform way to treat sentence pairs would be great.

In these specific cases, I would love to have an example of real input.