🔄 In a Training Loop
Daniel Gallagher
DanielGallagherIRE
AI & ML interests
NLP at the Institute for Applied Informatics in Leipzig, Germany.
Recent Activity
liked a model 29 minutes ago
jhu-clsp/mmBERT-base updated a model 2 days ago
DanielGallagherIRE/bert-base-cased-fineweb-nouns-only updated a model 2 days ago
DanielGallagherIRE/obfuscated-bert-fineweb-1B-originalOrganizations
Obfuscated ModernBERT
Models pretrained on various obfuscated versions of the same dataset to analyse how the obfuscation affects downstream performance.
CORAL Structural Obfuscation (OLD)
Encoder models trained on obfuscated versions of the fineweb-edu-1B dataset, as part of the InfAI CORAL project.
-
DanielGallagherIRE/obfuscated-bert-fineweb-1B-original
0.1B • Updated • 36 -
DanielGallagherIRE/obfuscated-bert-fineweb-1B-scramble-reverse-head
0.1B • Updated • 1 -
DanielGallagherIRE/obfuscated-bert-fineweb-1B-noun-pos
0.1B • Updated -
DanielGallagherIRE/obfuscated-bert-fineweb-1B-noun-propn
0.1B • Updated
Obfuscated FineWeb Edu
A collection of obfuscated version of a 20B-token sample of the FineWeb Edu dataset.
-
DanielGallagherIRE/fineweb-edu-20B-obfuscation
Viewer • Updated • 19.4M • 143 -
DanielGallagherIRE/FineWeb-Edu-20B-E1-POS-Removal
Viewer • Updated • 9.71M • 685 -
DanielGallagherIRE/FineWeb-Edu-20B-E1-Bag-of-Words
Preview • Updated • 888 -
DanielGallagherIRE/FineWeb-Edu-20B-E1-Mutual-Information
Viewer • Updated • 9.71M • 2.37k
CORAL Mutual Information Obfuscation (OLD)
Interesting Research
Obfuscated FineWeb Edu
A collection of obfuscated version of a 20B-token sample of the FineWeb Edu dataset.
-
DanielGallagherIRE/fineweb-edu-20B-obfuscation
Viewer • Updated • 19.4M • 143 -
DanielGallagherIRE/FineWeb-Edu-20B-E1-POS-Removal
Viewer • Updated • 9.71M • 685 -
DanielGallagherIRE/FineWeb-Edu-20B-E1-Bag-of-Words
Preview • Updated • 888 -
DanielGallagherIRE/FineWeb-Edu-20B-E1-Mutual-Information
Viewer • Updated • 9.71M • 2.37k
Obfuscated ModernBERT
Models pretrained on various obfuscated versions of the same dataset to analyse how the obfuscation affects downstream performance.
CORAL Mutual Information Obfuscation (OLD)
CORAL Structural Obfuscation (OLD)
Encoder models trained on obfuscated versions of the fineweb-edu-1B dataset, as part of the InfAI CORAL project.
-
DanielGallagherIRE/obfuscated-bert-fineweb-1B-original
0.1B • Updated • 36 -
DanielGallagherIRE/obfuscated-bert-fineweb-1B-scramble-reverse-head
0.1B • Updated • 1 -
DanielGallagherIRE/obfuscated-bert-fineweb-1B-noun-pos
0.1B • Updated -
DanielGallagherIRE/obfuscated-bert-fineweb-1B-noun-propn
0.1B • Updated