epsilon-compositionality

Code for our paper Measuring Idiomaticity in Text Embedding Models with 𝜀-compositionality
git clone https://esimon.eu/repos/epsilon-compositionality.git
Log | Files | Refs | Submodules | README

DateCommit messageAuthorFiles+-
2025-10-07 08:30Fix title/improve READMEÉtienne Simon1+3-4
2025-10-06 14:11model display name and factorizationÉtienne Simon1+52-43
2025-10-06 13:23add extended head-modifier tableÉtienne Simon1+15-0
2025-10-06 05:00add dataset statisticsÉtienne Simon1+29-0
2025-10-06 04:30Merge pull request #1 from SondreWold/cleanupÉtienne Simon14+216-1111
2025-10-06 04:28use distance epsilon and output latexÉtienne Simon6+216-206
2025-09-29 04:20remove unused codeÉtienne Simon8+0-905
2025-09-29 04:13rewrite everythingÉtienne Simon6+401-1
2025-09-29 04:12Move stuff around to prevent breakageÉtienne Simon5+441-422
2025-09-25 14:07ensure > 0 for simSondre Wold1+1-1
2025-09-25 14:07reproduce file updatesSondre Wold1+6-7
2025-09-25 07:54CleaningSondre Wold2+2-2
2025-09-25 07:51cleanup unused code i models.py and unused stuff from gitignoreSondre Wold2+3-89
2025-09-25 07:44add comments on normaltest, fix formattingSondre Wold1+15-8
2025-09-25 07:20Updates to data readme, remove normalization from experimentSondre Wold3+4-280
2025-09-25 07:15Refactor and clean up the codeSondre Wold7+522-856
2025-09-24 14:43add class to resultsSondre Wold2+13-3
2025-09-24 14:42add cleaned data with compositionality classSondre Wold1+273-0
2025-09-24 10:41updates, reproduce current tableSondre Wold5+262-52
2025-09-11 09:50fix broken urlSondre Wold1+2-1
2025-09-11 09:49Add modified datafile and descriptionSondre Wold2+293-0
2025-09-11 09:39Add script for end-to-end reproductionSondre Wold2+21-4
2025-09-11 09:38fix bug in get_epsilonSondre Wold1+1-1
2025-09-11 09:38add deberta and modernbert + a small effort validation of the diffSondre Wold1+58-2
2025-09-03 10:34rename test fileSondre Wold2+86-86
2025-09-03 09:26more statisticsSondre Wold1+24-7
2025-09-03 06:43plotting, random experiments, t-testSondre Wold3+255-2
2025-08-28 10:17get float from sent-transformersSondre Wold1+4-4
2025-08-27 12:32scripts for debuggingSondre Wold2+201-0
2025-08-27 12:31formatSondre Wold2+77-42
2025-08-26 07:55fast error handlingSondre Wold1+9-3
2025-08-26 07:00save distribution to json instead of csvSondre Wold1+29-18
2025-08-26 06:31changes to experiment class, save embeddingsSondre Wold2+31-30
2025-08-18 12:34no index on saved csvSondre Wold1+1-1
2025-08-18 12:31calculate epsilon for all result filesSondre Wold1+92-0
2025-08-18 12:00change hyphen to underscoreSondre Wold1+2-2
2025-08-18 11:09gitignoreSondre Wold1+2-0
2025-08-18 11:04update run fileSondre Wold1+5-0
2025-08-18 11:01make folder on saveSondre Wold2+24-40
2025-07-21 14:31fix bugSondre Wold1+3-1
2025-07-21 14:26remove word2vecSondre Wold1+0-35
2025-07-21 14:20update experiment classSondre Wold1+26-3
2025-07-16 11:38added more modelsSondre Wold1+185-10
2025-07-12 18:22add readmeSondre Wold1+1-2
2025-07-12 18:21add readmeSondre Wold1+2-0
2025-07-12 18:20add gitignoreSondre Wold1+210-0
2025-07-12 18:19add plot scriptSondre Wold2+40-2
2025-07-12 18:15remove old mainSondre Wold2+0-227
2025-07-12 18:13lintingSondre Wold1+65-46
2025-07-12 14:42Experiment classSondre Wold1+251-0
2025-07-03 13:49minimal common substring on SICKSondre Wold1+113-0
2025-06-13 12:51more filteringSondre Wold1+10-6
2025-06-12 14:11remove break statements, duhSondre Wold1+0-3
2025-06-12 13:51Add selection of device, change output file nameSondre Wold1+9-3
2025-06-12 13:43add main file from ÉtienneSondre Wold1+107-0