Skill UI

This guide outlines comprehensive best practices for ensuring the reproducibility of Machine Learning and Pattern Recognition research. It covers fixing random seeds, locking hardware/software dependencies (Python, CUDA, etc.), managing data splits, tracking pre-trained model versions, and structuring the required artifact package for external review. It is essential for building trust in published scientific results.

A comprehensive guide detailing best practices for achieving high reproducibility in recommender system research. It covers critical areas such as dataset version pinning, defining symmetric tuning protocols, selecting full-ranking metrics, handling non-determinism via multiple seeds, and structuring rigorous divergence analysis for academic publications.

This guide establishes the rigorous policy for submitting computational and empirical papers to academic journals. It mandates that authors provide sufficient data, source code, and detailed documentation (including execution order, random seeds, and OS used) to ensure that research findings can be fully reproduced by others, adhering to global standards like Option C.

A comprehensive guide for authors preparing methodological papers for submission to SMR. It outlines strict standards for creating a usable, reproducible package, including structured code, master scripts with fixed seeds, and formal data/code availability statements, ensuring the method can be run by others.

A comprehensive guide for ensuring the scientific reproducibility of research papers that rely on dynamic web data, such as web crawls, platform APIs, or live system traffic. It defines three reproducibility regimes (Frozen, Decaying, Unreplayable) and provides detailed instructions on documenting data decay, temporal splits, environment configurations, and computational seeds to maintain the rigor of published findings.

This guide outlines the mandatory requirements for authors publishing quantitative or mixed-method research in academic journals. It details the process of creating a complete replication package—including raw data, master scripts, seeds, and extensive README documentation—and depositing it in a trusted data repository (like Dataverse). It ensures scholarly rigor by guaranteeing that published findings are fully reproducible and traceable.

A comprehensive framework for ensuring reproducibility in Web Search and Data Mining (WSDM) research. It addresses unique challenges associated with behavioral data, covering data provenance, temporal splitting discipline, bias assumption reporting, statistical rigor (seeds, variance), and the 'honesty ladder' for non-reproducible results, helping researchers publish robust and verifiable work.

Stress-test core claims for Journal of Systems Engineering submissions: parameter sensitivity, initial conditions, alternative models and metrics, extreme scenarios, Monte Carlo repetitions, random seeds, software environments, data/code statements, and explicit failure boundaries.