# AI use cases for Polaris (ZJU-REAL/Polaris) Repo: [ZJU-REAL/Polaris](https://github.com/ZJU-REAL/Polaris) | Use cases: 4 | Page: https://dedupedai.com/use-cases/zju-real-polaris/ Polaris is a desktop app that runs the full AI research lifecycle locally, from literature survey and idea generation through GPU experiments, LaTeX paper writing, and paper review. This file holds every use case of this repo. Each one starts with a "## " heading and ends with its URL. - Build a cross-linked wiki for a research direction: https://dedupedai.com/use-cases/zju-real-polaris/build-literature-wiki-for-direction/ - Run a GPU experiment on your lab server with a Voyage: https://dedupedai.com/use-cases/zju-real-polaris/run-gpu-experiment-over-ssh/ - Fact-check a draft paper's citations and numbers: https://dedupedai.com/use-cases/zju-real-polaris/fact-check-draft-paper-citations-and-numbers/ - Query your research library from Claude Code or Cursor: https://dedupedai.com/use-cases/zju-real-polaris/query-research-library-from-coding-agent/ ## How to build a cross-linked wiki for a research direction with Polaris Repo: [ZJU-REAL/Polaris](https://github.com/ZJU-REAL/Polaris) | Difficulty: Intermediate | Time: 1 hour | Basis: From the README Ingest papers from OpenAlex, Semantic Scholar and arXiv into a wiki that is compiled up front. **Who it is for:** A PhD student or researcher starting a new topic who needs to survey a field quickly and keep the notes organized. **The problem:** Reading and linking dozens or hundreds of papers by hand takes weeks, and notes lose their connections over time. **The outcome:** A library of papers, each with a compiled wiki page (TL;DR, method, reusable ideas and concept backlinks), plus research digests you can sync to an Obsidian vault. ### Steps 1. Open the Polaris desktop app and create a direction library, then write its inclusion statement, goals and exclusions through the AI interview. 2. Add anchor papers so cold start can snowball citations from them into the library. 3. Let the Research Wiki ingest papers from OpenAlex, Semantic Scholar and arXiv and extract full text. 4. Open the compiled wiki pages and check the TL;DR, method and concept links for a few papers. 5. Turn on daily arXiv feed collection into your library and optionally sync the wiki to an Obsidian vault. ### Prompt to paste into your agent ```text Using the Polaris repo at ZJU-REAL/Polaris, set up a direction library for and run cold start from the anchor papers . Expected output: a count of ingested papers and a list of compiled wiki pages. Verify that each wiki page links to at least one concept backlink. ``` ### Caveat The README describes one shared wiki per paper and a concept is promoted only after two papers cite it, so early wiki pages will have fewer cross-links than a mature library. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/zju-real-polaris/build-literature-wiki-for-direction/ ## How to run a GPU experiment on your lab server with a Voyage with Polaris Repo: [ZJU-REAL/Polaris](https://github.com/ZJU-REAL/Polaris) | Difficulty: Advanced | Time: 2 hours | Basis: From the README Let an experiment Voyage plan, code, smoke-test and iterate over SSH under budget caps. **Who it is for:** A researcher with a lab GPU server who wants an agent to run a study while they keep control of remote writes and spending. **The problem:** Setting up experiments, watching logs and fixing failures across many runs takes constant manual attention. **The outcome:** A tracked experiment run with streamed logs, live metric curves, generated figures and an audit trail, with the agent asking you a question when it is stuck. ### Steps 1. Add your GPU server through the Experiment Lab with your SSH credentials, which Polaris stores Fernet-encrypted per user. 2. Start an experiment Voyage and answer the intake questions about the study. 3. Review the plan and the compute-budget check before approving any remote writes. 4. Watch the smoke test, the streamed logs and the live metric curves in the console. 5. Answer any question the run asks you, and review the generated figures and their VLM check when it finishes. ### Prompt to paste into your agent ```text Using the Polaris repo at ZJU-REAL/Polaris, start an experiment Voyage on the GPU server for the study . Expected output: a completed run with metrics, logs and figures. Verify that every remote write was approved through the gate and that the budget caps were not exceeded. ``` ### Caveat The README says the agent repairs failures under a time budget rather than a fixed retry count, so a stuck run will pause and ask you rather than finish on its own. Risk, credentials: The run uses your SSH credentials to reach a real GPU server, so only add a server you are authorized to use and review the command allow and deny lists before starting. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/zju-real-polaris/run-gpu-experiment-over-ssh/ ## How to fact-check a draft paper's citations and numbers with Polaris Repo: [ZJU-REAL/Polaris](https://github.com/ZJU-REAL/Polaris) | Difficulty: Intermediate | Time: 1 hour | Basis: From the README Verify each citation for existence and support, and check every number against the experiment record. **Who it is for:** A researcher preparing a submission who wants to catch fabricated references and mismatched results before sending the paper out. **The problem:** A draft can contain citations that do not exist, citations that do not support the claim, or numbers that differ from the experiment results. **The outcome:** A line-by-line citation report (existence and support labels), a number fact-check against the experiment record, and reviewer and meta-review comments. ### Steps 1. Open the Paper Writer project for your draft in Polaris so the references and experiment runs are linked. 2. Start Paper Review on the manuscript. 3. Read the citation report and check each line marked fabricated or unsupported. 4. Compare the flagged numbers with the experiment record and correct the draft. 5. Read the multi-perspective reviewer comments and the meta-review before deciding on submission. ### Prompt to paste into your agent ```text Using the Polaris repo at ZJU-REAL/Polaris, run Paper Review on the manuscript . Expected output: a citation report with existence and support labels and a number fact-check list. Verify that any citation marked fabricated is removed or fixed before the paper is marked as passing. ``` ### Caveat The README says citation support is checked against the library, Semantic Scholar and OpenAlex, so a paper that cites sources outside those will get a weaker check. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/zju-real-polaris/fact-check-draft-paper-citations-and-numbers/ ## How to query your research library from Claude Code or Cursor with Polaris Repo: [ZJU-REAL/Polaris](https://github.com/ZJU-REAL/Polaris) | Difficulty: Intermediate | Time: 45 minutes | Basis: From the README Expose Polaris's read-only tools as an MCP server so coding agents can search your knowledge base. **Who it is for:** A researcher who already works in Claude Code, Codex or Cursor and wants those agents to read their Polaris library while writing code or notes. **The problem:** Switching between the research app and a coding agent breaks flow, and the agent cannot see the papers and notes already collected. **The outcome:** A coding agent that can search and read your project's papers and wiki pages through the MCP server, without being able to change them. ### Steps 1. Open Polaris and find the MCP server settings, then check the self-check result. 2. Choose the transport the README lists, Streamable HTTP or stdio, that your agent supports. 3. Add the Polaris MCP server to Claude Code, Codex or Cursor using the connection details shown in the app. 4. Use the try-it playground in Polaris to confirm the tools return results for your project. 5. Ask your coding agent a question about your library and check the cited wiki pages. ### Prompt to paste into your agent ```text Using the Polaris repo at ZJU-REAL/Polaris, connect the MCP server for project to . Expected output: the agent lists papers from that project that match the query . Verify that the results come only from the chosen project and that no write tool was called. ``` ### Caveat The README says the MCP tools are strictly read-only and project-isolated, so the agent cannot edit your library or experiments through them. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/zju-real-polaris/query-research-library-from-coding-agent/