Skill UI

This comprehensive guide outlines the rigorous standards required for achieving experimental reproducibility in systems research papers (e.g., VLDB). It instructs authors on disclosing critical context, including hardware specifications, software configurations, data provenance, workload patterns, and handling measurement variance. Adhering to these guidelines ensures that all performance claims are fully traceable from raw data to published figures, meeting the highest standards of academic review.

A comprehensive guide for strengthening the reproducibility of academic papers, especially those involving systems claims (e.g., power usage, latency). It details maintaining a 'recipe ledger' for all reported metrics, ensuring constraint-aware measurement of environmental variables, and managing data/artifact consistency across the entire two-round revision process.

Provides comprehensive guidelines for packaging and submitting research artifacts (datasets, models, code) for major conferences. It details the difference between mere 'existence' (badge verification) and true academic 'reproducibility,' covering archival strategies, web data licensing, takedown resilience, and minimal viable package structure for long-term scholarly impact.

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.

A comprehensive guide for preparing replication and transparency packages for World Development manuscripts. It provides checklists and best practices for both quantitative (code reproducibility, data provenance, de-identification) and qualitative (evidence-to-claim trail, methods appendix) research. It addresses ethical constraints and the unique challenges of double-anonymized review.

This comprehensive guide outlines best practices for packaging research artifacts, including code, proprietary data logs, and models, for submission to top-tier academic venues. It addresses the crucial need for verifiability and reproducibility, providing strategies for handling limited space, proprietary datasets, and the absence of formal artifact grading systems, ensuring high academic credibility.

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.

Builds data-and-code replication packages for Management World (《管理世界》) after conditional acceptance. Generates Chinese READMEs, audits deposits, handles restricted commercial databases (CSMAR/Wind) and government micro-data, ensuring one-pass editorial reproducibility checks.

DataLad manages scientific datasets using Git and git-annex, handling large files efficiently by storing content remotely. It captures computational provenance for reproducible analysis, allowing users to clone, fetch, and publish datasets from repositories like OpenNeuro. Ideal for managing large data files and ensuring analysis reproducibility.