Paper Radar
Daily arXiv and bioRxiv radar where Jev answers one yes/no per plain-English interest for every new paper; fork-and-go GitHub Action.
Overview
Paper Radar reads every new arXiv and bioRxiv paper each morning and surfaces the few you should read. Instead of writing summaries, it asks Jev one yes/no question per plain-English interest for every paper and gets a probability back, such as 'introduces an agent benchmark, 0.97'; your repository's code applies thresholds and publishes a must-read page and RSS feed from a GitHub Action, with no server.
The README reports a real run reading 50 papers in 5 seconds for $0.002 (about 932 input tokens per paper), extrapolating to roughly $0.06 a day for all of arXiv; a list summary cites 501 papers judged in 33 seconds for $0.0196. It compares favourably with keyword alerts, embedding recommenders and LLM digest bots on reading every paper, explaining matches with per-interest probabilities, and offering tunable thresholds. A calibrate command measures calibration on your own labels, and an offline demo needs no key.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Yes / no | Does this paper match this interest (e.g. introduces an agent benchmark)? | yes · no |
How it works
- Fetch the day's new arXiv/bioRxiv papers.
- Ask Jev one noul per interest per paper.
- Apply thresholds in code and rank.
- Publish the page and RSS feed via GitHub Actions.
Strengths
- Reads all new papers cheaply.
- Per-interest probabilities explain matches.
- Calibrate command for threshold tuning.
Caveats
- Requires a TypeSafe key for live runs.
- Matches depend on how interests are phrased.
Getting started
Fork the repo, add your TypeSafe key as a secret and enable the GitHub Action; run paper-radar demo to see the page offline.At a glance
- Type
- App
- Status
- beta
- Language
- Python
- Runtime
- GitHub Actions, CLI
- Has code
- yes
- License
- MIT
- Stars
- 30
Use cases and integrations
How curated lists describe it
Research: scores every new arXiv and bioRxiv paper against plain-English interests with one Noul each and publishes the top picks as a daily page and RSS feed.
yibie/awesome-jevUnofficial daily arXiv and bioRxiv radar. Jev answers one Noul per plain-English interest for every new paper; code applies the thresholds and publishes a page and RSS feed from GitHub Actions. [Live demo](https://eliot5566.github.io/JEV-Paper-Radar/public/) needs no key.
AnotiaWang/awesome-decision-modelsIn GitHub Actions the links point at your own repo automatically (\`GITHUB\_REPOSITORY\`), so a fork needs no configuration. Locally, set \`output.feedback\_repo = "owner/name"\` or run \`paper-radar harvest --repo owner/name\`.
logicrw/awesome-jev-projectsLet Jev read every new arXiv paper each morning and surface the few you should read. Plain-English interests, calibrated probabilities, ~$0.06/day, fork and go.
wh000wh000/awesome-jev-liveDaily arXiv/bioRxiv radar that asks Jev one Noul per plain-English interest for every new paper, then publishes a must-read page and RSS feed from a fork-and-go GitHub Action; 501 papers judged in 33s for $0.0196, with a `calibrate` command that measures calibration on your own labels.
cobanov/awesome-jevLet Jev read every new arXiv paper each morning and surface the few you should read. Plain-English interests, calibrated probabilities, ~$0.06/day, fork and go.
hellogumbo/awesome-jevReads the morning's new arXiv papers against interests you write in plain English and surfaces the few worth opening.
MrJev/awesome-jev
Frequently asked questions
What does Jev decide?
Whether each new paper matches each of your plain-English interests.
How much does it cost?
About $0.06 a day for all of arXiv, per the README.
Do I need a server?
No, it runs as a GitHub Action and publishes a page and RSS feed.
Related projects
Local-first feed that merges arXiv, GitHub, Hacker News and Bilibili results, with optional Jev title-relevance scores as a filter.
ProjectbetaTypeScriptChrome extension that scores loaded X posts against your reading goal with Jev and ranks top matches pairwise.
ProjectbetaJavaScriptTech reading feed from Hacker News, Dev.to, Hashnode and Lobsters, ranked by Jev scores against your skill summary.
ProjectprototypeTypeScriptMulti-user Telegram bot that judges every new post in public channels with Jev templates and routes matches to your DM or channels.
ProjectbetaPythonTelegram channel X-ray: Jev judges up to 1,500 public posts for kind, paid ads, clickbait and pressure, producing a shareable card.
ProjectbetaJavaScriptLocal Telegram filter that batches channel messages to Jev or a local Laya model and keeps only those matching your conditions.
ProjectbetaPython