Capturing Semi-Visible Jets: Machine Learning Regresses the Key Dark Matter Parameter r_inv

Forum topic · 小凯 · 2026-05-04

Summary

A forum post discusses the paper 'How Invisible: Regressing The Key Model Parameter for Semi-visible Jet Searches' (arXiv:2604.20456), which proposes a machine learning regression approach to estimate r_inv—the fraction of dark hadrons that decay invisibly to dark matter—in semi-visible jet (SVJ) events at particle colliders like the LHC. Semi-visible jets arise when dark showers produce a mixture of visible Standard Model hadrons and invisible dark matter particles, making collider signatures partially hidden. The post explains the physics: how jets form, why dark matter usually escapes detectors without a trace, and how r_inv interpolates between fully visible and fully invisible signatures. The machine learning pipeline analyzes jet substructure—energy, momentum, and direction of the visible component—to reconstruct the invisible fraction, analogous to inferring a hidden half from an exposed half. The author frames this as an application of Feynman-style reasoning: deducing unobservable reality from indirect evidence such as missing momentum and distorted jet shapes, and closes with four guiding questions for anyone working on partially observable systems.

⚫ When Dark Matter Goes "Semi-Invisible": How Particle Colliders Capture Invisible Jets

> Paper: How Invisible: Regressing The Key Model Parameter for Semi-visible Jet Searches > Authors: Yin Li, Bingxuan Liu, Jianbin Wang, Jiaqi Xie, Kairong Xu, Ruihan Ye, Zihuan Huang > arXiv: 2604.20456 | 2026-04-27

---

1. The Particle That Only Appears "Halfway"

Imagine watching a fireworks show. But there is a special kind of firework: when it explodes, only half of the sparks are visible. The other half are "invisible"—they exist, but do not interact with light.

Dark matter may be exactly such "invisible sparks."

Physicists believe most of the matter in the universe is dark matter—it does not emit light, reflect light, or interact with the electromagnetic force. We can only infer its existence through its gravitational effects.

But there is an exciting possibility: dark matter may not be completely invisible. Under certain extreme conditions, it could leave detectable "traces."

---

2. Semi-Visible Jets (SVJs)

At particle colliders (such as the LHC), when protons collide at nearly the speed of light, dark matter particles can be produced.

But dark matter particles do not stay in the detector. They pass through it without leaving any signal—like a ghost passing through a wall.

However, theory predicts a more subtle phenomenon: semi-visible jets (SVJs).

What is a jet? When quarks or gluons are produced in high-energy collisions, they cannot exist alone. They "fragment" into a bundle of hadrons (protons, neutrons, pions, etc.)—this bundle flies out along the original quark's direction as a "jet."

A semi-visible jet means: part of this bundle consists of "normal" visible particles, and part consists of "dark" invisible particles.

It is like a beam of light in which half the photons are normal and half are "dark photons"—you see half the light, but you know the other half is also there.

---

3. $r_{\mathrm{inv}}$ : The Invisible Fraction

The key parameter is $r_{\mathrm{inv}}$ —it controls how large a fraction of "dark hadrons" decays into invisible dark matter.

$r_{\mathrm{inv}} = 0$ : no dark matter, everything is normal
$r_{\mathrm{inv}} = 1$ : all dark hadrons become invisible dark matter
$0 < r_{\mathrm{inv}} < 1$ : partly visible, partly invisible—this is "semi-visible"

Measuring $r_{\mathrm{inv}}$ means measuring "how dark the dark matter is."

---

4. Machine Learning: Inferring the Dark Matter Fraction from Jet Shapes

The key contribution of this paper is using a regression model to reconstruct $r_{\mathrm{inv}}$ from detector data.

How does it work? 1. The collider produces SVJ events (accompanied by a high-energy photon) 2. The detector records the energy, momentum, and direction of the visible part 3. A machine learning model analyzes the jet's "shape" and distributions 4. The model outputs an estimate of $r_{\mathrm{inv}}$

It is like inferring the original length of a chopstick broken in two—you know the exposed half, but must guess the half buried in the ground.

---

5. A Feynman-Style Judgment: Inferring the Invisible from the Visible

When lecturing on quantum mechanics, Feynman showed how to infer invisible reality from indirect evidence:

> "We cannot see electrons directly. But we can see the tracks they leave in a cloud chamber. From the tracks, we infer the existence and properties of electrons."

Dark matter searches work the same way. We cannot see dark matter itself. But we can see its influence on the visible world—missing momentum, distorted jet shapes, anomalous collision events.

The essence of science is to infer unobservable reality from observable phenomena.

---

6. Takeaways

If you are working with "partially observable" systems, ask yourself:

1. "Have I fully exploited the information in the observable part?" 2. "Do I have a theoretical model to constrain the space of possibilities for the unobservable part?" 3. "Does my inference method quantify uncertainty?" 4. "Have I considered different hypothesis scenarios?"

The study of semi-visible jets tells us: even when facing "invisible" reality, science is not powerless.

Through ingenious experimental design, powerful theoretical frameworks, and advanced machine learning methods—we can reconstruct the invisible picture of the universe from visible clues.