Uncovering simultaneous breakthroughs with a robust measure of disruptiveness An embedding-based disruption measure not only robustly captures disruptive works but also reveals simultaneous discoveries.

Uncovering simultaneous breakthroughs with a robust measure of disruptiveness | Science Advances www.science.org/doi/10.1126/...

Thanks!

📄 paper: www.science.org/doi/10.1126/...
💻 code: github.com/yy/embedding...
📦 pip: pypi.org/project/embe...
✍️ Blog: munjungkim.github.io/embedding-di...

w/ @skojaku.bsky.social @yyahn.bsky.social
Supported by @NSF TIP APTO, @AFOSR, and @nvidiabot.bsky.social

7/7

We flipped this around to systematically detect simultaneous discoveries across science, finding about 18,000 potential pairs. Manual checks on a sample of highly cited pairs confirmed about 80% were genuine.

6/7

Some findings:

- High-scoring papers on our metric are significantly more likely to be Nobel Prize-winning papers

- Papers where our metric and the disruption index diverge most kept turning out to be simultaneous discoveries — Higgs mechanism, reverse transcriptase, asymptotic freedom, ...

5/7

Building on the foundational idea of disruption index, w/ @skojaku.bsky.social & @yyahn.bsky.social we developed an embedding-based metric that preserves the core intuition while producing stable scores across simultaneous discoveries.

4/7

The Higgs mechanism is a perfect example. Three teams, same discovery — yet the disruption index gives one paper a top 2.8% score while the others fall in the bottom 0.1% and 0.04%.

3/7

The disruption index is a beautifully simple idea: if a paper is disruptive, future work cites it but stops citing its references. The stream of knowledge has bent.

But because it relies on local citation topology, simultaneous discoveries can throw it off.

2/7

Uncovering simultaneous breakthroughs with a robust measure of disruptiveness An embedding-based disruption measure not only robustly captures disruptive works but also reveals simultaneous discoveries.

The Higgs mechanism was proposed in 1964 by three independent teams.

But here is the puzzle🤔: the "disruption index" says Higgs's paper is among the least disruptive ever.

So what is going on?

In our new paper, just out in #ScienceAdvances, we take up this puzzle: doi.org/10.1126/scia... 👇

1/7

Research Scientist or Postdoctoral Research Associate - School of Data Science The School of Data Science at the University of Virginia and Professor Yong-Yeol (“YY”) Ahn, Quantitative Foundation Distinguished Professor, are excited to announce two professional research staff po...

🚨 A research staff position at my group at @uvadatascience.bsky.social!

I’m recruiting a research Scientist (or postdoc scholar). You can apply here! bit.ly/ylab-researc...

For more information: www.linkedin.com/feed/update/...

Inquiries: yyahn@virginia.edu with [job] in the subject line.

GitHub - skojaku/llm-quiz Contribute to skojaku/llm-quiz development by creating an account on GitHub.

Creating questions at a right level of challenge helps me learn nuances and go beyond surface-level understanding. But why not let students to do that?

I created a pedagogical tool that lets students to create challenging questions that stump LLMs.

Repo 👉: github.com/skojaku/llm-...