03RESEARCH MATERIALS
Research implementations and supporting artifacts, linked to their papers. Independent infrastructure documents the provenance of multilingual benchmarks.
Consult each repository’s workflow and limitations. Linking code does not establish independent reproducibility.
01CODE INDEX
Start with the paper, then follow its supporting code and documented methods.
Supporting research code
Code, per-item predictions, and analysis artifacts for studying cross-lingual consistency on parallel XNLI.
Related paper: Same Problem, Different Answer: Accuracy Hides Cross-Lingual Inconsistency in Multilingual Inference
License: MIT (code); CC BY 4.0 (outputs and documentation), with third-party exceptions
Supporting research code
Majority-vote ensemble sizing code, a frozen experimental protocol, and archived simulation outputs.
Related paper: How Many Advisors Are Enough? Predicting Useful Majority-Vote Ensemble Size under Correlated Decision Errors
License: MIT (code); CC BY 4.0 (archived research data)
Supporting research code
Prompt-format evaluation code, released correctness tensors, and Generalizability Theory analysis artifacts.
Related paper: How Many Prompts Certify a Ranking? A Generalizability Theory Protocol for Compute-Optimal, Dependable LLM Benchmark Evaluation
License: MIT (original code and documentation); third-party data and model terms apply
02CODE INDEX
Tools and metadata maintained beyond a single publication.
Reproducibility artifact
Machine-readable provenance metadata and an evidence-backed registry for multilingual and translated NLP benchmarks.
License: MIT (code); CC BY 4.0 (registry data and evidence snapshots), with third-party exceptions
↗CONTACT & COLLABORATION
I welcome conversations about research, graduate study, and collaborations in AI evaluation, multilingual NLP, and reliable AI systems.
ahsan.h.shihab@gmail.com