ArXiv cracks down on AI slop, may penalize careless endorsers
Endorsers who rubber-stamp low-quality AI papers could face account restrictions
ArXiv, the leading preprint repository for physics, math, and computer science, has long relied on an endorsement system to maintain quality: established academics vouch for newcomers based on direct collaboration or mentorship. This puts the endorser's own reputation on the line. Now, with a surge of AI-generated low-quality papers flooding the platform, ArXiv is actively banning accounts that upload such content—a welcome move for the research community. However, some argue that the problem starts upstream: endorsers who give out endorsements too freely, without vetting the submitter's work, enable the flood of slop.
One vocal researcher on Reddit proposes that ArXiv should not only ban the slop submitters but also retroactively warn endorsers who approved them. If an endorser is linked to multiple low-quality uploads (say, three strikes), they should face consequences—such as losing endorsement privileges temporarily or permanently. This would create accountability and incentivize more selective endorsement, preserving ArXiv's role as a trusted preprint server. Without such measures, the researcher warns, the entire endorsement system could lose meaning, and ArXiv risks becoming a dumping ground for AI-generated noise rather than a venue for genuine scientific progress.
- ArXiv's endorsement system requires researchers to vouch only for direct collaborators, but careless endorsements are enabling low-quality AI papers
- ArXiv has begun banning accounts that upload 'AI slop'—low-quality, likely machine-generated research
- Proposed solution: issue warnings to endorsers after the first violation, then penalize repeat offenders (e.g., three strikes) to restore accountability
Why It Matters
If ArXiv holds endorsers accountable, it could stem the tide of AI-generated junk and restore trust in preprint research.