Replication package for "Fine-Tuning Embeddings to Retrieve Related Threads on GitHub Discussions"

Replication package for a study on domain-adaptive representation learning for retrieving related threads on GitHub Discussions. It contains the source code, configuration files, datasets, fine-tuned model weights, and evaluation results needed to reproduce the reported experiments. Contents:- src/: fine-tuning (Triplet loss with online semi-hard mining), retrieval evaluation, query ablation, and significance testing code- config/: JSON configuration files for all experiments- data/datasets/: pairwise related/unrelated discussion data from five open-source repositories- outputs/: fine-tuned model, training metrics, grid-search summaries, rankings, and evaluation results- supplementary_material/: pipeline diagrams and example figures See README.md for installation and usage instructions. Code is released under the Apache License 2.0.