Explore hardware and electronic design configurations by repeatedly running programs, measuring results, and adjusting parameters. Track constraints and every trial, then report the best configuration found within the time and iteration limits.
Apply staged improvements to skill files after explicit user selection and a fresh review by a different model family. Each applied change includes a backup, review records, and provenance details for later inspection.
Review communications research by searching existing libraries and notes before academic databases. Produce a paper comparison table covering methods, results, and limitations, then summarize major approaches, evidence strength, and remaining research gaps.
Audit research experiments before using their results to support claims. An independent model checks reference data, metric calculations, result files, and evaluation scope, producing reports with evidence, findings, and specific actions to address concerns.
Build a paper outline from experiment results and review conclusions. Map claims to evidence, plan sections, figures, and citations, and allocate pages for the target venue, with another model reviewing the outline before it is finalized.
Create academic architecture diagrams and method illustrations with Gemini image generation. The workflow refines layout and visual style, then uses Claude to inspect arrows, labels, and readability and guide further revisions.
Search academic papers by topic, year, citation count, and free availability. Organize abstracts and publication details into a comparison table, and investigate author affiliations, funding sources, and citation links through OpenAlex.
Redirects the retired poster command to `/paper-poster-html`, preserving all supplied arguments. It helps users accustomed to the old command reach the replacement workflow; it no longer runs the former LaTeX poster process.
Understand a single arXiv paper through an AlphaXiv summary, with full text and LaTeX source available when needed. Use it to identify key findings or ask specific questions about methods, experiments, and equations.
Turn a broad research direction into ranked ideas with clear hypotheses and small experiments. Survey related papers, use another model to challenge candidates, and update priorities using pilot results when suitable computing resources are available.
Keep a local paper directory and an Overleaf project in sync through the Overleaf Git bridge. Review collaborators’ changes before importing them, check for conflicting edits, and push local revisions back for shared editing.
Improve research through up to four rounds of external language model review, fixes, and reassessment. Use an OpenAI-compatible API to identify weaknesses, prioritize changes, and keep a cumulative record of reviewer responses, actions, and experiment results.
Check running experiments, collect logs and result metrics, and summarize progress alongside baseline comparisons. It supports SSH servers, Vast.ai, and Modal, with optional Weights & Biases metrics and cost reporting for the supported compute services.
Explore a broad research direction through literature searches, idea generation, novelty checks, small pilot experiments, and critical review. Receive ranked ideas plus a refined proposal and experiment plan for the leading candidate, with evidence recorded across stages.
Compare machine learning experiment results from JSON or CSV files. Calculate improvements over a baseline, summarize variation across repeated runs, identify trends and unusual results, and suggest experiments to test possible explanations.
Turn research ideas and literature into a structured funding proposal. Supports KAKENHI and other funding programs, guiding literature positioning, research aims, methods, timelines, and budget justification, with external review to identify weaknesses in the draft.
Check paper numbers, comparisons, and evaluation scope against raw experiment results. A fresh reviewer from a different model family examines only source evidence, then produces reports identifying rounding issues, inconsistent settings, aggregation errors, and unsupported claims.
Analyze ARIS usage logs to propose evidence-based changes to skill instructions, reviewer prompts, and workflow defaults. Receive a report and focused patch proposals, with selected patches staged for a separate application process.
Turn a completed research paper into conference slides tailored to the talk format and duration. Produce a PDF, an editable PowerPoint deck, speaker notes, and a full rehearsal script with anticipated questions.
Turn descriptions of processes, systems, or schedules into Mermaid diagrams, which use text to define visuals. Save editable source, a Markdown preview, and a rendered PNG, then check syntax, arrow directions, labels, and readability.
Audit citations in a LaTeX paper against online sources. Verify that references exist, bibliographic details are accurate, and each cited work supports the surrounding claim, then produce reports with findings and proposed corrections.
Prepare a conference talk from a completed paper, including an outline, Beamer slides, editable PPTX, speaker notes, and a full script. Polish the visuals, check exported files, and optionally audit claims, citations, and anonymity.
Audit whether a paper's EVALUATION DESIGN actually measures what it claims and whether its reporting is complete — the validity layer family D (experiment-forensics) cannot reach. Three patterns: train/test leakage means the reported score may not measure generalization (HP-EVAL-LEAKAGE — adopts the Kapoor & Narayanan 8-type / 3-category leakage taxonomy; the illegitimate-proxy / sampling-bias / pretraining-contamination subtypes hand off as needs_external_check, naming but NEVER running Oren-2023 exchangeability / Shi-2023 Min-K% / Golchin-2023 Time-Travel / BIG-bench canary); a load-bearing LLM judge is conflicted (same model/family as a compared system) or unvalidated (no human-agreement, no bias control) (HP-JUDGE-VALIDITY); a declared condition/metric is dropped or switched to favor the method, or 'best' is chosen with no held-out set (HP-SELECTIVE-REPORTING). Verdict-bearing at L0/L1 from the DESCRIBED protocol — NOT repo-gated like experiment-forensics; L2 only CONFIRMS against split/preprocessing/result files. A fresh cross-model reviewer (gpt-5.6-sol xhigh, read-only, fresh thread per pass) PROPOSES findings, each span-anchored to a ledger claim_id; tools/adjudicate_findings.py DECIDES the verdict. Leakage and under-reporting are usually HONEST methodological errors — every finding describes a discrepancy to CHECK, never an accusation. An LLM generating GROUND-TRUTH labels is HP-FAKE-GT (experiment-forensics) — routed there, not here. Emits eval-design-forensics.findings.json; computes NO verdict. Detect-only. Triggers: "eval design audit", "evaluation validity", "train/test leakage", "data leakage", "is the score measuring generalization", "LLM judge bias", "is the judge validated", "selective reporting", "cherry-picked results", "评估设计审计", "评测有效性", "数据泄漏", "训练测试集泄漏", "裁判模型有没有验证", "选择性报告".
The English summary is being prepared. Showing the original description.
Synthesize the single strongest EVIDENCE-BOUND reviewer case to reject a paper, built ONLY from the evidence ledger (claims.json) + the other auditors' confirmed findings — never free-floating LLM critique. Two fresh cross-model codex threads: an attack writes the ~200-word rejection paragraph (every accusation tagged to an existing claim_id/finding_id), a defense decomposes it and rules each point against the anchored evidence. MEMO-ONLY: emits adversarial-case-builder.memo.md (fed to the adjudicator via --memo) and carries NO verdict weight — tools/adjudicate_findings.py lists it in ZERO_WEIGHT_SKILLS and caps it at info. Honest-null allowed (the paper may survive). Run LAST. Detect-only. Adapted from ARIS kill-argument. Triggers: "adversarial case", "strongest objection", "rejection memo", "kill argument", "最强拒稿点".
The English summary is being prepared. Showing the original description.
Is this your account? Link GitHub to bring it onto your creator page (Japanese)