SciGroveBeta
Physics

Towards anomaly detection searches for new physics signatures including Higgs bosons with weakly supervised machine learning

Chi Lung Cheng, Julia Gonski, Runze Li, Qibin Liu, Benjamin Nachman, Dennis Noll, Julie Khalilieh Romman, Liangyu Wu

Featured July 23, 2026

Original Paper

This analysis was generated by SciGrove. Upload your own PDFs or enter a DOI — and get the same AI breakdown on any paper.

Get started

AI-generated analysis — This is SciGrove's AI interpretation of the paper, not peer-reviewed content. Always refer to the original paper.

Simply

Scientists improved a special AI search called HAXAD that looks for new particles by spotting unusual patterns in Higgs boson events, making it much better at finding hidden clues than older methods.

In depth
The paper refines the Higgs And X Anomaly Detection (HAXAD) strategy for new physics searches at colliders. It introduces two novel embedding strategies (unsupervised VAE and semi-supervised contrastive encoder) to better represent collider event features, improving sensitivity to diverse signal models. Additionally, a new inference framework is developed to derive robust signal-agnostic and signal-specific cross-section limits, enhancing the statistical rigor of anomaly detection.

Key Takeaways

  • 1
    The study introduces two new ML-based embedding strategies (unsupervised VAE and semi-supervised contrastive encoder) that significantly improve the sensitivity of the HAXAD anomaly detection pipeline.
  • 2
    A novel inference framework is developed to derive both model-independent and model-dependent cross-section limits, providing a robust statistical interpretation for anomaly detection searches.
  • 3
    The enhanced HAXAD method consistently outperforms previous versions and traditional cut-based analyses across a wide range of Beyond the Standard Model (BSM) signal models, demonstrating its superior discovery potential.

Conceptual Flow

HIGH LEVEL
1
HAXAD: Finding New Physics

The HAXAD method uses AI to transform complex particle data, then estimates the normal background, and finally trains a smart classifier to find rare, unusual events that might be new physics.

Raw Particle Data
Transform and Analyze
Unusual Event Candidates
2
Improved Anomaly Detection

By using new AI ways to understand particle data, the updated HAXAD method finds new physics signals much more effectively than older techniques, setting tighter limits on where new particles could be.

Old Search Method
New Search Method
Compare Performance
Better Signal Finding