给还在做科研的朋友们的福音：
十年经验，一次调用。
香港科技大学的 𝐒𝐮𝐩𝐞𝐫𝐯𝐢𝐬𝐨𝐫-𝐒𝐤𝐢𝐥𝐥𝐬 把研究者十年的经验变成可用的 AI Skills，从想法到发表，全流程自动化学术助手。
Python 开源，已有 9,458 star。
以后做研究，可能先问 AI 导师，再问真导师。
HKUSTDial/Supervisor-Skills t.co
16 likes · 2 reposts · 953 views
Supervisor-Skills: Give Your AI a PhD Supervisor (9,631 Stars)
Your AI just got a PhD supervisor.
Supervisor-Skills is a trending GitHub repo (HKUSTDial/Supervisor-Skills, 9,631 stars and climbing) that distils a decade of PhD-supervisor research experience into installable AI skills. One prompt installs them into Claude Code, Cursor, or Codex.
The skills the repo ships with:
- Idea Evaluator: scores your research idea on a five-dimension framework
- Paper Writer: evidence-gated drafting with checked citations
- Pre-Submission Reviewer: a mock top-venue review before you submit
- Introduction Drafter, Figure Design Advisor, Deep Research, and a Vibe Research workflow
The skills also work in English (the README is Chinese-primary with an English mirror).
Repo: HKUSTDial/Supervisor-Skills
#AI #GitHub #ClaudeCode #Research #PhD #Productivity #Coding #TechNews
3 likes · 184 views
DocPilot: Interactive Agent Orchestration and Training for Long Visual Document Understanding
Multimodal long-document question answering requires models to seek sparse evidence across pages and modalities, understand heterogeneous document content, and reason over the collected evidence. Even advanced vision-language models (VLMs) can overlook sparse evidence and struggle with cross-modal reasoning in a single pass. Existing document agents further introduce multi-step interaction, but some still rely on predefined workflows or coarse-grained actions, limiting the flexibility and granularity of document reasoning. We introduce DocPilot, an agentic document framework that supports expressive orchestration and makes complex document solving learnable by small language models (SLMs). DocPilot orchestrates evidence seeking through page-level navigation and fine-grained evidence localization, and integrates local visual inspection, code-based numerical reasoning, and cross-modal reasoning for multimodal understanding. To transfer these capabilities to small models, we propose curriculum-based agentic training that progresses from single-ability learning to adaptive orchestration through supervision from data-grounded reasoning traces and executable interaction trajectories. Across MMLongBench-Doc, DocBench, FinRAGBench-V, and MADQA, DocPilot achieves strong gains with frontier VLM backbones, while DocPilot-4B delivers competitive performance against substantially larger models, demonstrating strong effectiveness and generalization across diverse long-document reasoning settings. The code is available at HKUSTDial/DocPilot.
Comments: 25 pages, 13 figures, 9 tables
🤯 港科大开源 Supervisor-Skills，把带博士十年的科研经验炼成可调用的 AI 技能：从想 Idea、查文献到实验、写论文、投稿。读研的人等于多了个副导师。
HKUSTDial/Supervisor-Skills
10 views
@GoSailGlobal 我最近用的是 paper-polish：
HKUSTDial/Supervisor-Skills
我拿自己的旧帖和书稿试了试，试完的结论是：用过一二十个去 AI 味的技能，它是最好的，没有之一。 t.co
90 views
Como é opensource, eu modifiquei para aceitar uns tipos de midias, e criei um bot exclusivo a minha esposa que é pesquisadora, usando HKUSTDial/Supervisor-Skills que se comporta como um orientador de PhD para auxiliar ela. 🧶
1 like · 11 views
DataVista: Diagnosing Multimodal LLMs on Data Video Understanding
Data video is a media form that integrates data visualization with video narrative, widely adopted in news reporting and business analysis. Compared with general video understanding, data video understanding places greater emphasis on accurately reading data from animated charts, integrating evidence across charts and time, and understanding how narrative organization and visual design communicate information. Yet existing benchmarks target either general videos or static charts, and data video understanding has not been systematically evaluated. We present DataVista, the first benchmark for data video understanding, containing 961 real-world data videos and 6,775 evaluation questions organized under a three-level progressive capability framework (data perception, temporal reasoning, narrative understanding) with 10 fine-grained question types across five topic domains. Systematic evaluation of 19 mainstream MLLMs shows that the best-performing model, Gemini-3.1-Pro, achieves 70.0% overall accuracy, still far below human expert performance, with models performing worst on Causal Reasoning and Narrative Structure. Increasing frame counts and adding subtitles mainly benefit data perception and temporal reasoning, with limited gains in narrative understanding. Further analysis of model responses identifies typical failure modes in chart reading, evidence judgment, and instruction understanding. The benchmark is available at HKUSTDial/DataVista.
Comments: 46 pages, 22 figures, 14 tables
最后
项目还在快速迭代（2026-07 已经出了 v2.1，新增 deep-research、paper-writer、paper-polish），GitHub 上点个 Star 支持一下，有问题可以去他们的微信交流群聊：
🔗 HKUSTDial/Supervisor-Skills
12/12
14 views
Mentions
给还在做科研的朋友们的福音： 十年经验，一次调用。 香港科技大学的 𝐒𝐮𝐩𝐞𝐫𝐯𝐢𝐬𝐨𝐫-𝐒𝐤𝐢𝐥𝐥𝐬 把研究者十年的经验变成可用的 AI Skills，从想法到发表，全流程自动化学术助手。 Python 开源，已有 9,458 star。 以后做研究，可能先问 AI 导师，再问真导师。 HKUSTDial/Supervisor-Skills t.co
Supervisor-Skills: Give Your AI a PhD Supervisor (9,631 Stars) Your AI just got a PhD supervisor. Supervisor-Skills is a trending GitHub repo (HKUSTDial/Supervisor-Skills, 9,631 stars and climbing) that distils a decade of PhD-supervisor research experience into installable AI skills. One prompt installs them into Claude Code, Cursor, or Codex. The skills the repo ships with: - Idea Evaluator: scores your research idea on a five-dimension framework - Paper Writer: evidence-gated drafting with checked citations - Pre-Submission Reviewer: a mock top-venue review before you submit - Introduction Drafter, Figure Design Advisor, Deep Research, and a Vibe Research workflow The skills also work in English (the README is Chinese-primary with an English mirror). Repo: HKUSTDial/Supervisor-Skills #AI #GitHub #ClaudeCode #Research #PhD #Productivity #Coding #TechNews
DocPilot: Interactive Agent Orchestration and Training for Long Visual Document Understanding Multimodal long-document question answering requires models to seek sparse evidence across pages and modalities, understand heterogeneous document content, and reason over the collected evidence. Even advanced vision-language models (VLMs) can overlook sparse evidence and struggle with cross-modal reasoning in a single pass. Existing document agents further introduce multi-step interaction, but some still rely on predefined workflows or coarse-grained actions, limiting the flexibility and granularity of document reasoning. We introduce DocPilot, an agentic document framework that supports expressive orchestration and makes complex document solving learnable by small language models (SLMs). DocPilot orchestrates evidence seeking through page-level navigation and fine-grained evidence localization, and integrates local visual inspection, code-based numerical reasoning, and cross-modal reasoning for multimodal understanding. To transfer these capabilities to small models, we propose curriculum-based agentic training that progresses from single-ability learning to adaptive orchestration through supervision from data-grounded reasoning traces and executable interaction trajectories. Across MMLongBench-Doc, DocBench, FinRAGBench-V, and MADQA, DocPilot achieves strong gains with frontier VLM backbones, while DocPilot-4B delivers competitive performance against substantially larger models, demonstrating strong effectiveness and generalization across diverse long-document reasoning settings. The code is available at HKUSTDial/DocPilot. Comments: 25 pages, 13 figures, 9 tables
🤯 港科大开源 Supervisor-Skills，把带博士十年的科研经验炼成可调用的 AI 技能：从想 Idea、查文献到实验、写论文、投稿。读研的人等于多了个副导师。 HKUSTDial/Supervisor-Skills
@GoSailGlobal 我最近用的是 paper-polish： HKUSTDial/Supervisor-Skills 我拿自己的旧帖和书稿试了试，试完的结论是：用过一二十个去 AI 味的技能，它是最好的，没有之一。 t.co
Como é opensource, eu modifiquei para aceitar uns tipos de midias, e criei um bot exclusivo a minha esposa que é pesquisadora, usando HKUSTDial/Supervisor-Skills que se comporta como um orientador de PhD para auxiliar ela. 🧶
DataVista: Diagnosing Multimodal LLMs on Data Video Understanding Data video is a media form that integrates data visualization with video narrative, widely adopted in news reporting and business analysis. Compared with general video understanding, data video understanding places greater emphasis on accurately reading data from animated charts, integrating evidence across charts and time, and understanding how narrative organization and visual design communicate information. Yet existing benchmarks target either general videos or static charts, and data video understanding has not been systematically evaluated. We present DataVista, the first benchmark for data video understanding, containing 961 real-world data videos and 6,775 evaluation questions organized under a three-level progressive capability framework (data perception, temporal reasoning, narrative understanding) with 10 fine-grained question types across five topic domains. Systematic evaluation of 19 mainstream MLLMs shows that the best-performing model, Gemini-3.1-Pro, achieves 70.0% overall accuracy, still far below human expert performance, with models performing worst on Causal Reasoning and Narrative Structure. Increasing frame counts and adding subtitles mainly benefit data perception and temporal reasoning, with limited gains in narrative understanding. Further analysis of model responses identifies typical failure modes in chart reading, evidence judgment, and instruction understanding. The benchmark is available at HKUSTDial/DataVista. Comments: 46 pages, 22 figures, 14 tables
最后 项目还在快速迭代（2026-07 已经出了 v2.1，新增 deep-research、paper-writer、paper-polish），GitHub 上点个 Star 支持一下，有问题可以去他们的微信交流群聊： 🔗 HKUSTDial/Supervisor-Skills 12/12