Skill UI

A comprehensive guide for executing and reporting robust statistical analyses for manuscripts in Global Change Biology (GCB). This skill ensures the analysis meets the high standards of reproducibility required for publicly archived data (DOI). It covers advanced methods including mixed/hierarchical modeling, time-series and spatial analysis, meta-analysis, and rigorous uncertainty quantification (parameter, structural, and scenario). It emphasizes statistical rigor over result fabrication, guiding the use of appropriate inferential machinery based on data structure.

A comprehensive guide for auditing and improving the scientific rigor of hardware architectural evaluations (e.g., HPCA style). It emphasizes establishing a clear 'fidelity contract' defining tools, scope, and configurations. It requires that all performance claims are backed by credible instruments, mandates capturing full machine state provenance, ensuring honest baseline tuning, and reporting results via per-workload distributions rather than single means.

This guide outlines the rigorous methodology for ensuring that computational research results, especially those involving hardware architecture and simulation, are fully reproducible. It emphasizes pinning the entire result chain—from simulator commits and configuration files to workloads, sampling recipes, and statistics scripts—and mandates creating a manifest for every figure. Key practices include recording machine state for silicon runs and containerizing the environment to prevent data drift.

This guide details the best practices for packaging, documenting, and releasing research artifacts (code, models, data) for top-tier ML conferences. It covers two stages: preparing a non-identifiable archive for anonymous review, and managing the public release runway, focusing on reproducibility, licensing, and sustained maintenance.

A comprehensive guide for academic authors detailing how to establish and maintain rigorous reproducibility in Machine Learning research papers (e.g., ICLR). It instructs on mapping every central claim to verifiable evidence, including recording seeds, variance, specific compute budgets (training vs. inference), data preprocessing scripts, and ethical considerations. The statement must function as a public contract for the scientific community.

This guide assists researchers in determining the optimal venue for their theoretical machine learning contributions. It provides a framework for evaluating whether a result is best suited for COLT, or if it belongs in other top-tier conferences (e.g., NeurIPS, ICML) or journals (e.g., JMLR). It emphasizes differentiating between rigorous mathematical proofs (bounds, separations) and purely empirical findings.

This guide helps researchers determine if their Machine Learning project is suitable for ICLR, or if a different top-tier conference (such as NeurIPS, ICML, or CVPR) is a better fit. It emphasizes the need for deep representation-learning insights, mechanistic understanding, and theoretical contributions over mere application performance.

This guide helps transform technically sound machine learning drafts into highly readable, conference-ready papers suitable for ICLR and OpenReview. It emphasizes structuring the core learning-representation insight upfront, clarifying contribution types (method, theory, benchmark), optimizing for public skimming, and providing clear reviewer navigation paths. Use this when ensuring maximum impact and clarity in a challenging academic review environment.

Provides comprehensive guidelines for packaging supplementary materials, including code, data, model weights, and benchmarks, for major machine learning conferences like ICML. It details the requirements for both the double-blind review stage (focusing on anonymization and minimal replication commands) and the final public release package (ensuring permanent public links and clear licensing).

A comprehensive guide for authors on drafting strategic, anonymous rebuttals and discussion replies for top-tier Machine Learning conferences like ICML. Learn how to triage reviewer objections by focusing on soundness, originality, significance, clarity, ethics, and reproducibility, adhering strictly to double-blind review constraints and maximizing impact in limited discussion rounds.

This guide provides a comprehensive framework for stress-testing Machine Learning experimental evidence before submission or rebuttal. It covers critical areas such as establishing strong baselines, conducting mechanism-isolating ablations, reporting variance, checking for data leakage, and disclosing compute costs. Use this to ensure your ML claims are scientifically sound, robust, and fair against rigorous academic scrutiny.

This guide provides structured methodologies for authors to position their research within the context of existing literature and concurrent submissions at top-tier machine learning conferences (e.g., ICML, NeurIPS). It emphasizes identifying the technical delta, addressing novelty concerns, and properly citing related work to maximize acceptance chances.