---
format: "aidr-story-markdown/v1"
id: "041d5fd8bbb761fb6e89fa51b15dcc34bce60ecdf1b107813a425cae65ca65de"
canonical_url: "https://aidr.today/041d5fd8?lang=en"
title: "AI Agents From 35 Labs Produce 936 Research Papers and Invent Peer Review"
lang: "en"
requested_lang: "en"
available_langs: ["en","vi"]
translation_fallback: null
fallback_fields: []
published_at: "2026-11-19T22:52:14.000Z"
category: "Agents"
topics: ["agent"]
source_urls: ["https://marketbrief.now/ai/ai-agents-from-35-labs-produce-936-research-papers-and-invent-peer-revie-ca96dc2d","https://huggingnews.com/ai/ai-agents-from-35-labs-produce-936-research-papers-and-invent-peer-revie-ca96dc2d"]
summary: "AlphaXiv has launched a preprint server specifically for AI agents after a three-month study in which nearly 1,000 autonomous agents from 35 labs produced 936 research papers. The agents, which included contributors from Stanford, MIT, and ETH Zurich, generated the manuscripts alongside their code, data, and model weights without human assistance in writing the text or code. During the study, the agents independently created their own journal to review each other's work, identified errors in published human research, and retracted their own mistakes. In blind reviews comparing the output to human papers on arXiv, the agent-authored work was judged to be more rigorous, though human authors were found to produce more original ideas."
---

# AI Agents From 35 Labs Produce 936 Research Papers and Invent Peer Review

> [Open the canonical story](<https://aidr.today/041d5fd8?lang=en>)

**Published:** 2026-11-19T22:52:14.000Z
**Category:** Agents
**Topics:** agent

## Summary

AlphaXiv has launched a preprint server specifically for AI agents after a three\-month study in which nearly 1,000 autonomous agents from 35 labs produced 936 research papers\. The agents, which included contributors from Stanford, MIT, and ETH Zurich, generated the manuscripts alongside their code, data, and model weights without human assistance in writing the text or code\. During the study, the agents independently created their own journal to review each other's work, identified errors in published human research, and retracted their own mistakes\. In blind reviews comparing the output to human papers on arXiv, the agent\-authored work was judged to be more rigorous, though human authors were found to produce more original ideas\.

## Sources

- [Story source](<https://marketbrief.now/ai/ai-agents-from-35-labs-produce-936-research-papers-and-invent-peer-revie-ca96dc2d>)
- [Story source](<https://huggingnews.com/ai/ai-agents-from-35-labs-produce-936-research-papers-and-invent-peer-revie-ca96dc2d>)

