Manage GPU compute jobs on Qizhi through a command-line tool. Find available nodes, submit distributed training runs, launch batches with different parameter combinations, and inspect, monitor, or stop jobs across workspaces.
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.
Get critical feedback on research ideas, papers, and experimental results from an external AI reviewer. Repeated discussions help identify unsupported claims, plan missing experiments, and document agreed improvements with priorities and estimated computing costs.
Investigate software errors, version conflicts, and API questions across official documentation, GitHub, technical Q&A, and Chinese developer communities. Results distinguish matching errors, source authority, and reported environments, giving you documented leads and next checks for debugging.
Develop patent application drafts from an invention description, covering prior art searches, claims that define protection scope, specifications, and examiner-style review. Compile documents for China, the United States, or Europe, with human checkpoints throughout.
Plan a 10–12 page systems research paper for conferences such as OSDI and SOSP. Get section page budgets, paragraph roles, and writing patterns that connect the research problem, design decisions, and evaluation evidence.
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.
Rent and manage vast.ai GPUs based on the training task rather than a chosen hardware model. The skill estimates requirements, compares offers by total expected cost, prepares the selected instance, and retrieves results before destroying it.
Fill unfinished research-wiki paper pages using summaries and abstracts from external sources. Organize each paper’s problem, method, results, limitations, and reusable ideas while preserving existing content by default and recording which source supplied the information.
Coordinate a research project from a broad topic through literature review, idea selection, experiments, and iterative review. Produce a research summary for writing, with optional paper drafting and PDF compilation, while recording stage progress for later resumption.
Get critical feedback from Claude on research ideas, papers, and experimental results. Through multiple review rounds, identify unsupported claims, missing experiments, and narrative weaknesses, then document agreed conclusions and a prioritized improvement plan.
Turns research reports and experimental results into a paper through outline planning, figure creation, LaTeX drafting, PDF compilation, and review. Produces editable source files and a compiled PDF, with checks connecting claims to evidence.
Find related papers across local PDFs, research notes, and web sources, then summarize their key contributions and relevance. Use existing collections as a starting point, remove duplicates, and search for literature that fills gaps.
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.
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.
Turn ARIS research reports in Markdown or JSON into a single HTML file for reading. Choose a report or dashboard layout, retain source version details, and optionally review whether the conversion preserves content and structure.
Explore computer architecture and circuit design settings by repeatedly running programs, measuring results, and adjusting parameters. Compare configurations against a baseline, respect design constraints, and receive a report with the best configuration found and the search history.
Check a paper's numerical consistency using fixed rules, then record blocking or warning decisions and outstanding findings. The workflow preserves a traceable correction history, while explicitly leaving broader semantic review unavailable in its Codex mode.
Organize paper reviews into individual concerns and draft responses within venue limits. Track evidence, unanswered issues, and approved commitments, then produce reviewer replies and a revision checklist that can be updated during follow-up rounds.
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.
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.
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.
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.
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.
Is this your account? Link GitHub to bring it onto your creator page (Japanese)