Lehigh AI Research Lab (LAIR)
Official@openlair · United States of America
Lehigh AI Research Lab provides advanced infrastructure for large-scale model training, mechanistic interpretability, and structured academic research synthesis.
Agent Skills by Lehigh AI Research Lab (LAIR)
Showing 173 vetted skills indexed across 1 GitHub repositories.
OpenLAIRaris-run-experiment
Deploys and runs ML training experiments on local, remote, Vast.ai, or Modal GPU resources.
OpenLAIRds-experiment
Executes and records auditable main research experiments against an accepted baseline.
OpenLAIRaris-ablation-planner
Designs and implements ML ablation studies from a reviewer perspective using Codex and Claude Code.
OpenLAIRaris-auto-paper-improvement-loop
Iteratively improves LaTeX research papers via GPT-5.4 review, fix implementation, and recompilation.
OpenLAIRds-optimize
Manages candidate briefs, optimization frontiers, and branch promotion for algorithm-first research quests.
OpenLAIRautoresearch
Runs autonomous modify-verify-keep iteration loops for code improvement, debugging, security audits, and shipping.
OpenLAIRaris-mermaid-diagram
Generate Mermaid diagrams from requirements with syntax verification and rendered PNG review.
OpenLAIRaris-paper-compile
Compiles LaTeX papers to PDF, auto-fixes errors, and verifies submission readiness.
OpenLAIRaris-result-to-claim
Evaluates experiment results against intended claims and routes to next research action.
OpenLAIRaris-pixel-art
Generate pixel art SVG illustrations for READMEs, docs, and slides.
OpenLAIRaris-idea-discovery
Orchestrates a literature-to-validated-idea pipeline with novelty checks, pilot experiments, and reviewer feedback.
OpenLAIRaris-idea-creator
Generates, validates, and ranks research ideas with literature surveys and pilot experiments.
OpenLAIRds-write
Drafts evidence-grounded research papers and reports with LaTeX venue templates and citation integrity checks.
OpenLAIRaris-auto-review-loop
Automates iterative research review cycles using Codex MCP until positive assessment.
OpenLAIRaris-research-refine-pipeline
Chains method refinement and experiment planning into one end-to-end research proposal pipeline.
OpenLAIRaris-vast-gpu
Provisions, manages, and destroys vast.ai GPU instances based on training task requirements.
OpenLAIRds-full-pipeline
Orchestrates an end-to-end autonomous research pipeline from literature scouting to paper finalization.
OpenLAIRaris-feishu-notify
Send push or interactive notifications to Feishu/Lark via webhook or bridge.
OpenLAIRaris-experiment-plan
Generates claim-driven experiment roadmaps with ablation matrices, run orders, and compute budgets for research papers.
OpenLAIRaris-paper-plan
Generate a structured section-by-section paper outline from review conclusions and experiment results.
OpenLAIRds-baseline
Establishes verified research baselines through attach, import, reproduce, or repair routes.
OpenLAIRaris-paper-poster
Generate conference posters from compiled LaTeX papers as PDF, PPTX, and SVG outputs.
OpenLAIRds-review
Audits research paper drafts and produces skeptical review reports, revision logs, and experiment TODO lists.
OpenLAIRds-scout
Frames research tasks by scouting literature, evaluation contracts, and baseline candidates.
Frequently Asked Questions About Lehigh AI Research Lab (LAIR)
FAQPage SchemaWhat specific research tasks are supported by these capabilities?▼
The lab supports end-to-end research lifecycles, including literature review synthesis, grant proposal drafting, manuscript formatting, and automated rebuttal planning for conference submissions.
Which technical personas benefit from these research resources?▼
These resources are designed for machine learning researchers, computational scientists, and academic engineers requiring structured pipelines for model training, interpretability analysis, and scholarly documentation.
What are the primary dependencies for running these training pipelines?▼
Most training and orchestration modules require PyTorch, CUDA-enabled GPU clusters, and specific distributed backends like DeepSpeed, Megatron-Core, or Ray for multi-node scaling.