GAINER: Gain-Aware Allocation of Transformer Depth for Efficient Cross-Encoder Re-Ranking (code and results)

Code and registered runs of the paper GAINER: Gain-Aware Allocation of Transformer Depth for Efficient Cross-Encoder Re-Ranking (I. Mohd Jamadi and S. Anuar, preprint, 2026, doi:10.5281/zenodo.23340638; submitted to the International Journal of Advanced Computer Science and Applications). Same contents as GitHub release v1.0.0 of https://github.com/ibrahimjamadi98/GAINER.

GAINER-v1.0.0-code.zip: the two Google Colab notebooks, saved with the outputs of their registered runs: GAINER_v3.ipynb (main run, hypotheses H1–H3, pre-registration SHA-256 e77241073ba22c17b36151adcffacaf3e7f47ff36b3fd38cbebae413026b4746) and GAINER_v3_1.ipynb (second registered family, H4–H6, SHA-256 1f4050281d4dbf8a27543aa6672f7295ea86880db50d1aecea2818ec52ef11b1), with README, licence and requirements.

GAINER-v1.0.0-results-v3.zip, GAINER-v1.0.0-results-v3_1.zip: for each run, the pre-registration file, the run manifest, per-query results for every backbone, seed, held-out BEIR dataset, method and budget, measured latency, and all tables and figures.

No dataset is redistributed; the BEIR datasets are downloaded from their public releases.