GitHub · Where software is built

Labels Milestones New issue

fix(tokenization): replace fragile substring matching with character offsets from annotations

#264

· hanneshapke opened

on Mar 2, 2026

fix(data): shuffle samples before train/val split to prevent ordered-batch bias

#263

· hanneshapke opened

on Mar 2, 2026

feat(training): enable class weights computed from label frequency to handle class imbalance

#262

· hanneshapke opened

on Mar 2, 2026

feat(eval): switch to seqeval for entity-level F1 metrics

#261

· hanneshapke opened

on Mar 2, 2026

feat(data): mix in real-world annotated PII datasets to improve distributional variety

#260

· hanneshapke opened

on Mar 2, 2026

feat(model): disable coreference resolution task and focus training on PII detection

#259

· hanneshapke opened

on Mar 2, 2026

feat(model): evaluate upgrading from DistilBERT to a stronger base encoder

#258

· hanneshapke opened

on Mar 2, 2026

feat(model): use deeper MLP classification head instead of single linear layer

#257

· hanneshapke opened

on Mar 2, 2026

feat(model): add CRF layer on top of token classifier for valid BIO sequence decoding

#256

· hanneshapke opened

on Mar 2, 2026

When changing providers we should refresh the app input boxes or keep the providers frozen

#137

· Davidnet opened

on Jan 29, 2026

Extend Dataset with Increased Diversity and Complexity

#37

· hanneshapke opened

on Nov 28, 2025

Create Demo Video for Yaak Proxy

#35

· hanneshapke opened

on Nov 27, 2025