Abstracted is a free tool that reads about 2,400 new arXiv papers a week and orders them around what you actually work on.
Build my research feedSign up with Google
Pick a profile to see how the same day of research reads for different people.
You work on
3 papers worth your attention today
Researchers investigate how AI agents behave when faced with impossible tasks, specifically analyzing how peer influence and social propagation lead to boundary-crossing and unauthorized system modifications in multi-agent environments.
StanfordMODA is a post-training reinforcement learning framework that mitigates mode collapse in LLMs by using multi-agent role conditioning and a prompt-adaptive quality gating mechanism to jointly optimize for generation quality and output diversity.
Researchers quantify the 'English-Forcing Tax' in multi-agent LLM systems by comparing native-language pipelines against English-mediated ones, demonstrating that forcing inter-agent communication through English significantly degrades performance across typologically diverse languages.
A sample profile, ranked against this week's real papers — roughly 2,400 of them. You read three.
Every paper is already public. The hard part is knowing which handful is yours.
~2,400
new papers a week
Across eight arXiv categories in AI, ML, vision, language, robotics and signal processing.
Every one
read and briefed
Each paper gets a short plain-language brief, keywords, and its key figure pulled out.
~20
papers you actually see
Ordered by how close the work is to your research profile, how recent it is, and how strong it looks.
Nothing is hidden from you. A keyword filter silently drops everything it doesn't match, and you never learn what you missed. Abstracted orders instead of deletes — the work closest to yours comes first, and the rest is still there when you go looking.
01
Pick your topics once. We turn them into a vector profile of your field — not a list of keywords to string-match against.
02
Every weekday we pull new papers across eight arXiv categories and write a short brief for each one, with keywords and key figures extracted.
03
Closest to your work first, weighted by how recent and how strong each paper looks. Papers you have already scrolled past do not come back, and a daily nudge arrives at 8am in your own timezone.
| Scholar Alerts | Abstracted | |
|---|---|---|
| What it matches on | The query terms you set up | The meaning of your research profile, as a vector |
| What arrives | An email per alert, per query | One feed, strongest match first |
| Short brief on every paper | Title and a snippet | A plain-language brief, written for each paper |
| Key figure pulled out | Extracted automatically from the PDF | |
| Already-seen papers stop reappearing | One paper can arrive from several alerts |
We were researchers trying to keep up with our own fields, and the effort had quietly spread everywhere: alert emails, saved arXiv searches, a newsletter or two, papers screenshotted from social media, and a browser full of tabs we were definitely going to read.
None of it was broken, exactly. It just meant every morning started with triage instead of reading, and the papers that mattered most were indistinguishable from the ones that merely matched a keyword.
So we built the research feed we wanted for ourselves, and then made it free for everyone else who has the same morning.
“For researchers, by researchers”
Abstracted installs like a native app on iOS and Android, and keeps your last feed readable offline.
| Follow specific researchers | Separate alert per author | Boosted inside the same feed |
| Cost | Free | Free |
This compares what each tool is built to do, not an exhaustive feature audit. Scholar Alerts apply their own ranking and personalization — the difference is that an alert is optimized for a query you defined, and Abstracted is optimized for the field you work in. Keep the alert for the search you must not miss.
