Yang Ruofeng

Turn a feature request into a working prototype, starting with a minimal path from input to output and adding features through individual checks. Record consequential assumptions before coding, then use a separate model to find decisions missing from that record.

中文简介正在准备中。当前显示英语简介。

• Building a working prototype from a feature description
• Extending a prototype one checked feature at a time
• Recording assumptions made during implementation

Turn a vague research approach into a concrete method proposal with focused contributions and minimal validation experiments. Ground the plan in relevant papers, then refine it through GPT-6-Astra reviews while preserving the original problem and resource constraints.

中文简介正在准备中。当前显示英语简介。

• Turn a rough research idea into an implementable method
• Simplify a proposal with too many components
• Clarify novelty against related research

Run model training, fine-tuning, inference, and batch jobs on Modal’s cloud GPUs. Estimate hardware needs and cost, generate Python launch scripts, run or deploy workloads, monitor logs, and retrieve results without managing remote servers.

中文简介正在准备中。当前显示英语简介。

• Test CUDA code without a local GPU
• Run model training or LoRA fine-tuning on cloud GPUs
• Deploy a model inference API or vLLM service

Draft a patent specification from claims and invention disclosure materials. The skill prepares sections from background to detailed embodiments and abstract, then checks whether the description supports each element of the claims.

中文简介正在准备中。当前显示英语简介。

• Drafting a patent specification from existing claims
• Checking written support for every claim element
• Preparing drafts for CN, US, or EP requirements

Draft a research paper section by section in LaTeX, a document typesetting format, using an outline and research materials. Apply venue formatting, incorporate figures, verify bibliography records, and check that repeated theoretical statements agree.

中文简介正在准备中。当前显示英语简介。

• Drafting paper sections from an outline and research findings
• Applying conference or IEEE paper formatting
• Checking bibliography metadata and removing uncited entries

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.

中文简介正在准备中。当前显示英语简介。

• Turning a research report into a paper draft
• Creating paper figures and tables from experiment data
• Preparing a manuscript for a target venue

Review patent claims and specifications with an AI acting as a patent examiner. It identifies clarity, novelty, and disclosure issues, applies revisions, and conducts a second review, producing a report of changes and remaining concerns.

中文简介正在准备中。当前显示英语简介。

• Check claims and specifications before filing
• Find unclear terms and missing technical explanations
• Review novelty against known prior art

Manage mathematical proof work across separate runs, preserving sources, established results, and unresolved steps. Attempt and audit proofs locally, prepare focused manual GPT Pro handoffs when blocked, and check returned answers before producing a clearer final proof.

中文简介正在准备中。当前显示英语简介。

• Continuing proof work from a previous run
• Preparing a focused GPT Pro handoff
• Checking a returned proof against local sources

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.

中文简介正在准备中。当前显示英语简介。

• Find related work for a research topic
• Summarize relevant papers already stored locally
• Combine Zotero highlights with Obsidian research notes

Search patent databases and academic literature for earlier disclosures related to an invention. The workflow extracts technical features, compares overlapping approaches, and produces a report on relevant references, potential patent coverage, and areas where novelty may be strongest.

中文简介正在准备中。当前显示英语简介。

• Find patents and papers related to an invention
• Compare technical features with existing disclosures
• Identify potential overlap with patent claims

Turn a rough research approach into a focused plan with concrete implementation details and a small set of validation experiments. Repeated Claude reviews help refine the method while keeping the original problem and resource constraints in view.

中文简介正在准备中。当前显示英语简介。

• Turning a rough research idea into an implementable method
• Narrowing a proposal to one main contribution
• Choosing minimal experiments to test core claims

Turn a research paper into a conference poster using its figures and findings. Produce one HTML/CSS file and a print-ready PDF, with measured layout checks and Gemini reviews of content accuracy and final appearance.

中文简介正在准备中。当前显示英语简介。

• Create a conference poster from a research paper
• Redesign an existing research poster
• Check poster claims against the source paper

Plan and run ablation studies, experiments that remove or replace parts of a method to test their contribution. Organize reviewer questions, priorities, expected outcomes, and compute costs, then record results and findings.

中文简介正在准备中。当前显示英语简介。

• Test how much each component contributes
• Plan experiments that address reviewer questions
• Prioritize comparisons within a compute budget

Polish existing academic PPTX or Beamer slides through page-by-page Codex review against a reference PDF. Adjust typography, spacing, colors, and text overflow while preserving content, keeping the original deck, and recording each change.

中文简介正在准备中。当前显示英语简介。

• Improve slide readability before an academic presentation
• Match PPTX visual balance to a Beamer reference
• Fix text overflow and layout drift across slides

Turn a research report and experiment results into an English paper with figures, LaTeX source, and a compiled PDF. Coordinate planning, drafting, Gemini reviews, and two revision rounds, preserving PDFs and an improvement log.

中文简介正在准备中。当前显示英语简介。

• Turning a research report into a paper
• Drafting a paper from experiment results
• Preparing a paper for ICLR, NeurIPS, or ICML

Develop a vague research approach into a concrete method proposal through repeated Gemini reviews. Keep the original problem and constraints in focus while specifying implementation details, removing unnecessary components, and outlining the minimum experiments needed to test the claims.

中文简介正在准备中。当前显示英语简介。

• Turning a vague research approach into a concrete method
• Keeping revisions aligned with the original problem
• Simplifying an overly complex research proposal

Iteratively improve research by having Claude Code review code, results, and drafts while Codex implements fixes. Repeat until the assessment meets the stopping criteria or the round limit is reached, preserving feedback and unresolved issues.

中文简介正在准备中。当前显示英语简介。

• Repeating research reviews and fixes before submission
• Checking whether claims match code and experiment results
• Tracking how reviewer concerns are addressed across rounds

Turn a broad research direction into ranked ideas, a refined proposal, and an experiment plan. The workflow combines literature searches, novelty checks, small pilot experiments, and Gemini review to assess evidence and identify useful next steps.

中文简介正在准备中。当前显示英语简介。

• Explore concrete ideas within a broad research topic
• Check how proposed ideas differ from prior work
• Compare promising ideas through small pilot experiments

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.

中文简介正在准备中。当前显示英语简介。

• Checking claims before writing a paper
• Interpreting ambiguous experiment results
• Planning experiments to fill evidence gaps

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.

中文简介正在准备中。当前显示英语简介。

• Check loss and evaluation trends during model training
• Detect NaN values and diverging loss
• Review flat or noisy training metrics

Develop robotics research ideas through literature review, feasibility filtering, novelty checks, and external critique. Get ranked proposals with specific benchmarks, comparison methods, failure metrics, and simulation-first pilot plans matched to available infrastructure.

中文简介正在准备中。当前显示英语简介。

• Exploring research topics in manipulation or navigation
• Choosing robotics ideas that fit available infrastructure
• Planning pilot studies with simulation or existing logs

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.

中文简介正在准备中。当前显示英语简介。

• Check whether remote experiments have finished
• Compare collected metrics with the matching baseline
• Investigate NaN values or diverging training runs

Build a PDF from a LaTeX paper, diagnose compilation errors, and repair reference or citation problems. Check the resulting layout, page count, anonymity, and embedded fonts, then report fixes and remaining warnings.

中文简介正在准备中。当前显示英语简介。

• Build a PDF from LaTeX source files
• Fix compilation, reference, and citation errors
• Check figure and table layout

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.

中文简介正在准备中。当前显示英语简介。

• Illustrating a model architecture for a paper
• Turning a method description into a workflow diagram
• Reviewing diagram arrows and labels