# On some AI research topics, more than half of all papers are from 2026

Canonical: https://tiptreesystems.com/blog/half-of-all-papers-on-some-ai-topics-are-from-2026

2026-10-29 · Team Tiptree

More than half of all papers ever written on some AI research topics were published in 2026.

We looked at which research areas were growing fastest, busiest, or cooling off on Lacuna.

The 3 fastest-emerging research areas:  
- Group-relative RL (GRPO variants): 66% of papers published in 2026  
- Memory poisoning in AI agents: 53%  
- Entropy collapse in RL for language models: 51%

Two of the five youngest areas focus on RL training. Two focus on attacks against AI agents.

Meanwhile, just 2% of papers on overcoming local receptive fields in vision models appeared this year. Federated learning was at 3%.

The contrast raises a bigger question: Are we seeing AI research priorities shift in real time, or are these topics simply going through short-lived hype cycles?

Which emerging research area do you think will still matter in 3 years?

*First posted on [LinkedIn](https://www.linkedin.com/feed/update/urn:li:activity:7521573349945376768), October 29, 2026.*
