Results for legal-bert-small-uncased

[atreyasha]

Hi @iliaschalkidis,

Posting the results of legal-bert-small-uncased as per #8:

The validation and test results per dataset (task) in a table format, similar to the one presented in the main README.md file.

Test

Dataset	ECtHR Task A	ECtHR Task B	SCOTUS	EUR-LEX	LEDGAR	UNFAIR-ToS	CaseHOLD
Model	μ-F1 / m-F1	μ-F1 / m-F1	μ-F1 / m-F1	μ-F1 / m-F1	μ-F1 / m-F1	μ-F1 / m-F1	μ-F1 / m-F1
Legal-BERT-Small (Chalkidis et al., 2020)	70.4 / 62.6	71.3 / 69.4	71.3 / 59.7	66.1 / 48.2	87.8 / 82.0	97.4 / 81.7	72.9 / 72.9

Validation

Dataset	ECtHR Task A	ECtHR Task B	SCOTUS	EUR-LEX	LEDGAR	UNFAIR-ToS	CaseHOLD
Model	μ-F1 / m-F1	μ-F1 / m-F1	μ-F1 / m-F1	μ-F1 / m-F1	μ-F1 / m-F1	μ-F1 / m-F1	μ-F1 / m-F1
Legal-BERT-Small (Chalkidis et al., 2020)	72.8 / 65.9	73.7 / 73.3	75.6 / 68.5	73.4 / 54.4	87.8 / 81.4	97.1 / 76.3	74.7 / 74.7

PS: I am also running other lighter models such as DistilBERT and ALBERT, so will also get back to you on these.

The generated log files all_results.json for all five seeds and datasets (tasks). You can simply attach those in your message (post). To do so, you can simply drag and drop the files (or better a .zip file with all of them).

results.tar.gz

I'll rapidly review the results and then update the leaderboard with your reported results with a reference to your name (@atreyasha) or another identifier (full name, article) that you may find more appropriate (if this is ok with you of course)

We don't have a paper as yet, for now just a reference to my GitHub account would be great :)

[iliaschalkidis]

Posting the results of legal-bert-small-uncased as per #8:

Thank you @atreyasha for the detailed results and log files.

PS: I am also running other lighter models such as DistilBERT and ALBERT, so will also get back to you on these.

That's great! I'm looking forward 😃