Skill UI

Orchestrates PyTorch-native TorchTitan 4D parallelism (FSDP2, TP, PP, CP) for large-scale LLM pretraining, scaling from 8 to 512+ GPUs with Float8, torch.compile, and distributed checkpointing to boot Llama 3.1, DeepSeek V3, or custom models.

Hypogenic automates LLM-driven hypothesis generation and testing over tabular datasets, blending literature insights with empirical analysis to accelerate discovery in domains like deception detection or AI content evaluation.

Handles IMO/Putnam/USAMO/AIME style problems by generating refined proofs, then running a fresh-context adversarial verifier to catch flaws, finally producing honest confidence reports and verified PDFs when possible.

AKF is the native metadata format specifically designed for tracking AI-generated or AI-modified content. It allows users to stamp files (including DOCX, PDF, images, and code) with detailed trust scores, source provenance, and compliance metadata. This tool is essential for professional workflows requiring strict auditing, compliance assurance, and verifiable data lineage, especially concerning regulations like the EU AI Act, HIPAA, and SOX.

Master advanced patterns for composing LangGraph subgraphs, which is essential for building robust, multi-agent systems like planner-executors or hierarchical teams. This skill addresses critical issues such as shared state contract propagation (P21) and callback scoping (P28). Learn to manage state across independent graph components using sophisticated dispatch patterns: compiled nodes, Send fan-out, and Command bubble-up.

This powerful compiler ingests heterogeneous research inputs, including PDF papers, code repositories, experiment logs, and raw notes. It transforms this diverse data into a comprehensive, structured Agent-Native Research Artifact (ARA). The system extracts core knowledge atoms, builds cognitive maps (claims, concepts, heuristics), and generates a verifiable, executable knowledge package for advanced AI consumption and research iteration.

This skill runs an automated evolution process designed to systematically optimize the seven core policy surfaces of an AI agent harness (e.g., planner, contextBuilder, reviewer). By mutating these policies and scoring the variants in a sandbox, it identifies crucial performance improvements without requiring expensive foundation model retraining. It is ideal for empirical configuration tuning and establishing optimized baselines.

This tool executes the advanced `metaharness/darwin` security benchmark (ADR-155). It rigorously evaluates the security detection capabilities of AI systems against a controlled ground-truth corpus, providing critical metrics like TPR, FPR, and patch-pass rates. It compares the champion model against multiple baselines, offering empirical data crucial for iterative security improvement and drift detection.

Provides comprehensive guidelines for designing, auditing, and documenting empirical evidence for high-level AI research submissions (e.g., AAAI). It covers essential methodologies including strong baselines, single-factor ablations, robustness testing (shift, seed, prompt), statistical significance, and ensuring perfect alignment with reproducibility checklists and ethical reporting. Use this guide before submission to ensure empirical evidence strongly supports all claimed AI contributions.

A comprehensive guide detailing best practices for designing and auditing empirical evaluations of Large Language Models (LLMs). It addresses critical pitfalls such as data contamination, baseline unfairness, model version drift, and decoding parameter sensitivity, ensuring that published results are robust, reproducible, and scientifically sound for academic publication.

A detailed guide for authors submitting theoretical machine learning papers (like COLT) regarding the appropriate inclusion and design of numerical simulations. It emphasizes that numerics should only visualize proved bounds, phase transitions, or separations, maintaining the focus on theoretical proofs rather than empirical data. Provides rules for plotting, reporting variability, and maintaining academic honesty.

This skill guides authors on how to position, structure, and strengthen empirical Natural Language Processing (NLP) manuscripts for submission to the Conference on Empirical Methods in Natural Language Processing (EMNLP). It provides deep insights into conference fit, required evidence (baselines, error analysis, datasets), proper framing, and anticipating reviewer expectations for top-tier AI venues.