Congqiao Li

Working on @CMSexperiment at @PKU1898

[Peking University @PKU1898]

Huge congratulations to PKU alumni Wang Hong and Deng Yu, recipients of the 2026 Fields Medal — mathematics’ highest honor! 🧮

Announced at the 2026 International Congress of Mathematicians (ICM) in Philadelphia, the pair are among four mathematicians worldwide to claim this prestigious prize, awarded every four years to outstanding scholars under 40.

The awards are historic on several levels.

—First time two mathematicians who completed undergraduate studies in mainland China win the Fields Medal in the same year.

—Both joined Peking University as undergraduates in 2007.

—Wang Hong becomes only the third female Fields Medalist ever, the first Chinese woman, and the first recipient with a bachelor’s degree from a mainland Chinese university.

We are immensely proud of their extraordinary breakthroughs and lasting contributions to global mathematics.

[Congqiao Li @congqiao_li]

"Anyone knows adam?"

[INSPIRE HEP @inspirehep]

Diederik P. Kingma and Jimmy Ba's 2015 conference paper
"Adam: A Method for Stochastic Optimization"
https://t.co/CtevRQCfKi
reaches 4,000 citations.

[CMS Experiment CERN @CMSExperiment]

Teaching CMS to trace particles! 🧑‍🏫

For the first time, CMS is using machine learning to reconstruct collisions at the LHC – making it faster and even more precise than the traditional approach particle flow with #MachineLearning.

🗞️Read more: https://t.co/FgcfJx7gdX

[Congqiao Li @congqiao_li]

In practice, we believe the background should be even lower than 30 - thanks to
1) new Run-3 HH→4b trigger strategy, and
2) unsaturated scaling capability of the jet-free network.
Let's wait and see what the LHC will ultimately deliver😉

[Congqiao Li @congqiao_li]

After 4 months of further iterations, we refined the sensitivity for HH→4b 😀
With existing ATLAS or CMS data (Run 2+3), ~20 HH signal events are achievable with <30 background. Let’s see it realized at the LHC ✨
Check out our updated paper: https://t.co/18wRbP1iqA

[Google DeepMind @GoogleDeepMind]

We’re helping to unlock the mysteries of the universe with AI. 🌌

Our novel Deep Loop Shaping method
published in @ScienceMagazine could help astronomers observe more events like collisions and mergers of black holes in greater detail, and gather more data about rare space phenomena. 🧵

[Congqiao Li @congqiao_li]

Higgs pair revolution paper out! 🚀 https://t.co/18wRbP0KB2
Calibration method is still evolving. We'd love your feedback! (on alphaXiv)

Also check out Tianyi's ML4Jets talk: https://t.co/Apw49lPYgU

[Congqiao Li @congqiao_li]

I used to feel uneasy that "supervised" models like GloParT or Sophon for HEP weren't seen as "foundation models"... Then I came across Microsoft's "foundation model for Earth system" pretrained with MAE loss. That finally brought me peace of mind. Much appreciated!😇

[nature @Nature]

Nature research paper: A foundation model for the Earth system

https://t.co/wQyLSgHPo6

[Congqiao Li @congqiao_li]

This confirms it: AI is a game-changer for HEP analyses. The next leap in sensitivity will come from AI engineering. I wrote this in my PhD outlook - it felt bold then, but now it feels so real.
Huge thanks to the organizers! Isola d’Elba is an incredible place for a workshop.

[Congqiao Li @congqiao_li]

Honored to present at Higgs Pairs Workshop 2025! Our latest work (w/ Tianyi) reveals the astonishing potential of HH→4b. Traditional jet-based methods are hitting a ceiling, but full event-level classification boosts sensitivity by >500%📈!!! Begun, the new HH chapter has. [1/4]

[Congqiao Li @congqiao_li]

With LHC data ➡️2025, we’re already able to astonishingly approach a ~3σ HH observation - and there’s still room to grow😲 Stay tuned for our arXiv paper. Unreal, considering HL-LHC projections across all channels were ~4.5σ just months ago.
Slides: https://t.co/HzXVZOpST2
[2/4]

[Congqiao Li @congqiao_li]

A boosted topology review generated by DeepSearch. The level of detail is insane...

https://t.co/OW8MtZkxmm

[Congqiao Li @congqiao_li]

2. At the core of LHC analyses lies a classification problem. Use the tagger well—cut on it, fit the mass, and let resonant new physics emerge.
3. Use the latent space well. The gift from a monster tagger is the complete jet repr it offers. New opportunities are awaiting us.

[Congqiao Li @congqiao_li]

This marks a milestone in our 3 years of research: a general-purpose jet tagger and mass regression model is now part of CMS. We call it the Global Particle Transformer (GloParT).
Please guide us deeper into SM and, if fortune favours, present us with new physics.🙏

[Congqiao Li @congqiao_li]

We are accelerating resonance discoveries at the LHC with Sophon⚛️! (thanks #3bodyproblems)
Sophon:
🔍 Defines optimal discriminants, pushing widespread resonance search sensitivity to the frontier.
🤖 Enhances model-agnostic searches via high-powered transfer learning.

[Congqiao Li @congqiao_li]

The philosophy is revealed in the Sophon paper: https://t.co/CTceARsR0M
Some main points are:
1. Make the jet tagger a monster. Don’t be afraid to add more classes. It won’t hurt your performance, but instead, it helps.
2.

[Congqiao Li @congqiao_li]

🫡🫡

[CMS Experiment CERN @CMSExperiment]

Years of dedication, precision, and consistent teamwork! ⚖️The CMS experiment's W boson mass measurement was a marathon, not a sprint.
🗞️Discover the scientific and personal journey of the W mass team, as told in their own words: https://t.co/TVklw3HN8S

[CERN @CERN]

Happy 70th anniversary CERN 🎉

#OnThisDay 70 years ago, twelve European countries came together to establish what has become the world’s largest #ParticlePhysics laboratory, home to the #LHC, and where the #HiggsBoson was discovered 12 years ago.

In how many ways can you say #HappyBirthday?

In this video, representing our 24 Member States, CERN people share their birthday wishes in their native language.

#CERN70

[Javier M. Duarte @jmgduarte]

Our search for Higgs boson pairs in the HH->bbVV->bb4q final state is finally public just in time for #ICHEP2024! This is @rkansal47's @UCSDPhysics PhD thesis work with many other @CMSExperiment collaborators @crissms21 @congqiao_li @Caltech_LHC @Fermilab

[Congqiao Li @congqiao_li]

Now on arXiv: https://t.co/CTceARsR0M
Nice collaboration with @jmgduarte, Dr. Huilin Qu, @GregorKasieczka, @rkansal47, @crissms21, and more!
From inside the experiment to outside applications, we are bringing this technology real and vivid.✨✨

[Congqiao Li @congqiao_li]

We are accelerating resonance discoveries at the LHC with Sophon⚛️! (thanks #3bodyproblems)
Sophon:
🔍 Defines optimal discriminants, pushing widespread resonance search sensitivity to the frontier.
🤖 Enhances model-agnostic searches via high-powered transfer learning.