Check BibTeX references against web records and export field differences in a spreadsheet. Independent checking depends on an isolated web agent; without isolation, results flag potential issues rather than certify matching entries.
Compare local BibTeX references with independently retrieved web records to identify missing or reordered authors, incorrect years, venues, and identifiers. Export a side-by-side spreadsheet with per-reference verdicts and evidence for discrepancies.
Develop mathematical proofs in Lean, a system that checks formal proofs, continue unfinished projects, or audit existing results. Check that the theorem matches the original claim, examine its assumptions, and prepare sources and commands for reproduction.
Create model architecture diagrams and method illustrations for academic papers using Gemini. Codex plans the figure and reviews generated images for correct connections, content, layout, and readability, guiding repeated revisions toward a clearer final figure.
Review mathematical proofs in LaTeX for logical gaps, missing assumptions, and unjustified steps. Attempt counterexamples, develop explicit repairs, and repeat the review, producing an audit report that distinguishes completed fixes from assumptions and unresolved proof obligations.
Build a section-by-section paper outline from experiment results and review findings. Map claims to evidence, allocate pages for the target venue, and plan figures and citations in PAPER_PLAN.md before drafting the paper.
Build a runnable prototype from a plain feature request, then add capabilities one step at a time. The skill records consequential assumptions before coding, checks each step, and reports remaining placeholders, deferred work, and review limits.
Turn a paper outline and research narrative into a section-by-section LaTeX draft. Incorporate figures and verified bibliography entries, adapt formatting to the target venue, and review clarity, evidence alignment, and consistency across sections.
Reviews research ideas, papers, and experimental results through repeated critical discussion. Identifies unsupported claims, missing experiments, and weaknesses in the paper’s story, then records agreed conclusions, evidence requirements, and prioritized improvements in a review document.
Monitor machine learning training through WandB or training logs. Check for invalid numbers, worsening metrics, and stalled improvement, then report whether to continue, wait, or stop, preserving evidence and notes when stopping a run.
Manage mathematical proof projects across Codex runs, keeping sources, attempts, and unresolved steps organized. Prepare manual GPT Pro handoffs for difficult steps, then check returned arguments and edit the proof for clarity.
Apply staged skill improvements selected by the user after a fresh review of the actual changes. It backs up target files, applies passing patches, and records review evidence and change history for later inspection or reversal.
Build a PDF from a LaTeX paper, diagnose compilation errors, and fix common problems with citations, references, and figures. Check page limits, anonymity, and embedded fonts, then report remaining warnings before submission.
Evaluate whether completed experiments support a research claim using a separate Codex reviewer. Identify unsupported conclusions and missing evidence, then determine whether to revise the claim, run additional experiments, or move toward paper writing.
Create paper figures and LaTeX comparison tables from experiment data in JSON or CSV files. Save reproducible plotting scripts, use consistent styling, and check rendered figures for readability, fair comparisons, and captions supported by the data.
Launch machine learning training jobs on local or remote GPUs. The skill checks available GPU memory, synchronizes code, saves logs, and verifies startup, with optional experiment tracking and support for Vast.ai and Modal.
Check a paper’s numbers, comparisons, and evaluation scope against raw experiment files. A fresh reviewer receives only source evidence and paper files, then reports discrepancies in rounding, aggregation, settings, and reported improvements.
Turn requirements into Mermaid diagrams, including process flows, system structures, and timelines. Save editable diagram code, a Markdown version, and a PNG preview, then check syntax, arrow directions, labels, and readability.
Monitor model training through WandB metrics to detect invalid numbers, worsening loss, and declining evaluation results. Ambiguous trends receive a Codex review, while stopped runs retain key metrics and the reason for stopping.
Check mathematical proofs in LaTeX with review by a separate AI model. Identify missing assumptions and unsupported steps, explore counterexamples, repair derivations, and keep an audit record of issues, changes, and follow-up reviews.
Improve a compiled LaTeX paper through two rounds of Claude review, edits, and recompilation. The workflow addresses inconsistent reasoning, unsupported claims, unclear writing, and visual presentation, while keeping a cumulative log of changes.
Improve research through repeated Gemini reviews, code changes, experiments, and analysis. The workflow records full feedback and actions, then stops when both the score and readiness criteria are met or the round limit is reached.
Have Gemini critically review research developed with Codex. Multiple rounds of discussion examine logical gaps, missing experiments, and narrative weaknesses, then turn the findings into experiment plans, paper outlines, and documented research priorities.
Evaluate whether completed experiments support a research claim by checking cited evidence and requesting a Codex review. Identify unsupported conclusions, missing comparisons, and needed revisions, then determine whether to add experiments, change direction, or prepare for writing.
Is this your account? Link GitHub to bring it onto your creator page (Japanese)