Choose timezone
Your profile timezone:
18th International Workshop on Boosted Object Phenomenology, Reconstruction, Measurements, and Searches at Colliders
→
Europe/Zurich
Krupnicza 33, 31-123 Kraków
Andrzej Konrad Siodmok
(Jagiellonian University (PL))
Description
Overview
BOOST 2026 is the 18th conference of a series of successful joint theory/experiment workshops that bring together the world's leading experts in theoretical and experimental collider physics to discuss the latest progress and develop new approaches on the reconstruction of and use of jet substructure to study Quantum Chromodynamics (QCD) and search for physics beyond the Standard Model.
Note: BOOST 2026 will be held in the week just after PSR 2026 in Manchester. For attendees of both, Manchester and Kraków are well-connected by direct flights.
Warning:
We received information that some of the participants were contacted by someone offering help with hotel reservations. We would like to inform you that the organizers do not call or send emails offering such services and that we strongly discourage using them.
| Previous editions: | International Advisory Committee: | |
|
Local Organizing Committee:
- A. Siodmok (UJ, chair)
- P. Bruckman de Renstrom (IFJ PAN, ATLAS)
- I. Grabowska-Bold (AGH, ATLAS)
- A. Kusina (IFJ PAN, Theory)
- L. Motyka (UJ, Theory)
- J. Otwinowski (IFJ PAN, ALICE)
- J. Pajorska (UJ, Secretary)
- W. Placzek (UJ, Theory)
- R. Poncelet (IFJ PAN, Theory)
- E. Richter-Was (UJ, ATLAS)
- S. Sapeta (IFJ PAN, Theory)
- A. van Hameren (IFJ PAN, Theory)
- J. Whitehead (UJ, Theory)
Sponsors:
Patronage:
Contact
Participants
-
-
Registration
-
BOOST Camp
-
1
BOOST camp (experiment)Speaker: Nurfikri Norjoharuddeen (Lappeenranta-Lahti University of Technology (FI))
- 2
-
1
-
10:30
Coffee Break
-
Welcome to Krakow!
- 3
-
4
Introduction (theory)Speaker: Simone Caletti (Università degli Studi di Torino, INFN Torino and ETH Zurich)
-
12:45
Lunch break
-
Jet reconstruction, calibration and performance
-
5
Progress in reconstructing, classifying and calibrating hadronic objects in ATLAS
Hadronic objects reconstruction, classification & calibration are key ingredients of many physics analysis in ATLAS. The collaboration is continuously improving their performance by refining various aspects of the related procedures such as statistical methods, data-driven approach or cutting-edge machine learning techniques. This contribution presents highlights of the work and its impact of the jet and missing transverse momentum reconstruction performance.
Speaker: Fang-Ying Tsai (Stony Brook University (US)) -
6
Jets and missing transverse momentum performance in Run 3 of the CMS experiment
The LHC managed to deliver more than 200/fb of integrated luminosity to ATLAS + CMS in 2024 and 2025, and with this exceeded the expectations for Run 3. Data were collected with an average of 60 pileup interactions happening at the same bunch crossing, and thus challenging the CMS detector and jet reconstruction. We present the latest developments in pileup mitigation, jet calibration, and jet + missing transverse energy performance for LHC Run 3.
Speaker: Theodoros Chatzistavrou (Helsinki Institute of Physics(FI)) -
7
Particle Flow reconstruction for CMS phase 2
Ongoing developments of the particle-flow reconstruction for the Phase-2 upgrade of the CMS experiment are presented, with emphasis on algorithmic changes relevant for extreme pileup conditions. The contribution focuses on the integration of new detector inputs, in particular high-granularity calorimetry and precision timing, and their use in particle reconstruction and pileup mitigation. Developments in linking strategies, ambiguity resolution, and object building are discussed.
Speaker: Andreas Hinzmann (Deutsches Elektronen-Synchrotron (DE)) -
8
Machine Learning Approaches to the Calibration of the Signals and their Classification as Pile-up in the ATLAS Calorimeters
Employing machine-learned local calibrations for the basic calorimeter signals in the ATLAS experiment at the Large Hadron Collider (LHC), which are formed by clustering topologically connected cell signals (topo-clusters), shows indications of significant performance improvements in terms of accuracy and precision. The most successfully trained model so far is a dense neural network (DNN) employing a heteroscedastic loss function and implementing a Gaussian mixture in a probabilistic approach. This model is found to provide the best performing calibration in the highly stochastic signal environment dominated by the pile-up that is characteristic for the proton–proton collisions at the LHC, indicating a significant mitigation of these pile-up contributions and the associated fluctuations. In addition, independent classification networks have been studied with the goal to reduce the effect of pile-up on reconstructed calorimeter jets by reweighting the topo-cluster signal contribution to the jet kinematics. The corresponding weighting functions are based on the amount of signal arising from pile-up in a given topo-cluster. This talk will present the status of the latest developments for both calibration and classification, as well as a brief outlook on future applications in ATLAS.
Speaker: Mr Buddhadeb Mondal (Czech Academy of Sciences (CZ))
-
5
-
15:20
Coffee Break
-
Energy flow and correlators
-
9
One point charge correlator in DELPHI Open Data
The one-point charge correlator probes the charge-asymmetric angular structure of hadronic final states in $e^+e^-$ collisions. In perturbative QCD, the structure is intimately linked to the axial-vector-vector triangle anomaly, making it highly robust against higher-order corrections and offering a novel, theoretically clean perspective on the forward-backward asymmetry, $A_{\rm FB}$. We present the first measurement of this observable using DELPHI Open Data collected at $\sqrt{s} = 91.2$ GeV. The data, corrected for detector effects, exhibit a clear $\sin(2\theta)$ modulation, in agreement with both the PYTHIA 8.3 Monte Carlo prediction and theory calculation at N$^3$LO in QCD, representing the first direct observation of the differential asymmetry shape in data. This work provides experimental insights into charge correlator measurements and motivates future precision electroweak programs at the FCC-ee or through further analysis of archival LEP data.
Speaker: Jingyu Zhang (Vanderbilt University (US)) -
10
Resolved four-point energy-energy correlators: phenomenology, and measurement in Z+Jets events at CMS
The immense flux of high-momentum jets produced by the LHC provides a unique opportunity to study the detailed dynamics of high-energy QCD. So-called “energy-energy correlator” observables provide a particularly compelling opportunity to uncover these dynamics by probing the internal structure of jets while maintaining a close connection to fundamental ingredients of QCD theory. The bulk scaling properties of 2- and 3-point energy-energy correlators have already been measured by several experiments, but the resolved structure of these observables is still completely unknown. We will discuss the phenomenology of resolved four-point energy-energy correlators, including new experimentally-viable observables and first-principles calculations. We will then proceed to discuss recent measurements of these observables by the CMS experiment in Z+jets events at a center-of-mass energy of 13 TeV. We will further discuss the practical challenges in measuring and unfolding these high-dimensional, highly-correlated data. These results provide a stringent test of parton shower dynamics and QCD factorizations, and provide a basis for new insights into the high-energy dynamics of QCD.
Speaker: Simon Rothman (Massachusetts Inst. of Technology (US))
-
9
-
Triggers and real-time inference
-
11
Ultra-Fast Hadronic Tagging at 40 MHz: Machine Learning for the CMS Phase-2 Level-1 Trigger
We present advances in hadronic object tagging for the Phase 2 Upgrade of the CMS Level 1 Trigger. With advances in Fast Machine Learning; more powerful FPGA processors; and the introduction of track and particle flow reconstruction at the Level 1 Trigger, jet tagging will become feasible for the first time in this system. We show the implementation of tiny jet taggers using a DeepSets architecture with only O(1000) parameters. Our models are capable of processing 1 billion jets per second and with a latency of around 200 ns. Furthermore, we will present the physics performance enhancements that jet tagging can bring to the Level 1 Trigger. For the first time, boosted resonances are reconstructed as large-radius jets with dedicated algorithms, enabling trigger selection based on jet mass and substructure. In addition, a multi-class jet tagger on small-radius jets allows us to select physics targets such as di-Higgs events with looser kinematic thresholds. These results represent a significant step towards a more “physics-aware” trigger system capable of targeting and preserving challenging signals despite the high-luminosity environment.
Speaker: Santeri Laurila (CERN & Helsinki Institute of Physics (FI)) -
12
Real-time jet substructure tagging at the ATLAS hardware trigger
We present a convolutional neural network (CNN) for real-time jet substructure identification in the ATLAS hardware trigger at the High-Luminosity LHC (HL-LHC). This approach introduces substructure discrimination at Level-1, enabling efficient separation of jets from hadronically decaying boosted objects (e.g. W/Z/H bosons and top quarks) from the dominant QCD background under extreme pileup conditions. The CNN operates on jet “images” built from NxN grids of trigger towers or calorimeter cells centered on jet seeds.
We evaluate performance using inputs from both the existing Global Feature Extractor (gFEX) and the future Global Trigger systems, and demonstrate robustness across multiple jet definitions, pileup mitigation strategies, and calibration schemes currently under development for HL-LHC. Hardware feasibility is assessed through detailed studies of latency and resource usage on FPGAs and emerging AI-accelerator architectures (e.g. AMD Xilinx AI Engines).
This study demonstrates the deployability of deep-learning-based jet substructure tagging at the hardware trigger level. The resulting rate reduction for selections requiring both jet pT and substructure enables significantly lower thresholds for large-radius jets, substantially improving sensitivity to boosted hadronic signatures.Speaker: Tianjia Du (University of Chicago (US))
-
11
- 13
-
Poster session
-
14
Unlocking Hadronic W/Z Boson Polarization Measurements via decorrelation methods
W and Z boson polarization is a direct probe of electroweak symmetry breaking, as the longitudinal states are physical remnants of the Higgs mechanism and their production fraction is uniquely sensitive to anomalous gauge couplings and new physics.
While leptonic decay channels of W/Z provide clean polarization measurements through the lepton angular distribution, they suffer from limited branching fractions, neutrino reconstruction ambiguities, and restricted kinematic reach. The hadronic channel — dominant in branching fraction and fully reconstructible in the boosted regime — has remained experimentally inaccessible for polarization studies due to the overwhelming QCD multijet background, which will be a more promiment effect in HL-LHC.Due to the large QCD background it is inevitable that we must need a tagger to get sufficient signal purity. However, on the other hand the main challenge with pure polarized state classification taggers (beside QCD background) is that they suffer from a fundamental limitation: a tagger trained to discriminate longitudinal from transverse polarization states necessarily learns the same angular correlations one wishes to measure, creating a circular dependency that biases any subsequent shape analysis. We argue instead that a better approach is a QCD-vs-signal discriminant that is explicitly decorrelated from the polarization-sensitive observable (calling X), preserving the angular information in the data for unbiased extraction.
We present a three-dimensional extension of the Designed Decorrelated Tagger (DDT) [1], which simultaneously decorrelates an existing tagger score from the jet soft drop mass and also a cos θ* proxy observable (X) constructed in the large-R jet rest frame. This construction ensures a flat QCD selection efficiency at a fixed working point, making possible the use of mass sidebands for data-driven background estimation in different regions of X. We implement and test multiple X definitions based on subjet momenta obtained from declustering the jet.
Using the public JetClass [2] dataset and existing taggers, we demonstrate that the 3D DDT approach on these taggers successfully makes the decorrelation for the QCD background, and paves the way to create a workable analysis strategy blueprint for future hadronic W/Z polarization measurements.
1: JHEP 05 (2016) 156
2: https://zenodo.org/records/6619768Speaker: Amartya Rej (Technische Universitaet Dortmund (DE)) -
15
Preference-Optimized Generative Models for Underconstrained Inference in Particle Physics
Precise reconstruction of event kinematics in the presence of invisible particles constitutes a fundamental underconstrained inference problem in particle physics. In processes such as dileptonic $t\bar{t}$ production, multiple undetected neutrinos lead to a multimodal solution space, where several kinematically consistent configurations can explain a single observed event.
We present a preference-optimized generative framework for underconstrained inference, built on the event-level foundation model EveNet. Our approach augments a diffusion-based generative model with Direct Group Preference Optimization (DGPO), a post-training strategy that shifts the learned distribution toward physically preferred solutions while preserving its multimodal structure.
Evaluated on the $t\bar{t}$ dilepton channel, the preference-optimized model improves reconstruction fidelity and reduces unfolded uncertainties compared to both the baseline EveNet diffusion model and $\nu^2$-Flows across key observables. These results establish preference optimization as an effective paradigm for generative inference in underconstrained systems, with direct implications for precision measurements at the LHC.
Speaker: Yi-Ren Wu (National Taiwan University (TW)) -
16
Classifying hadronic objects in ATLAS with ML/AI algorithms
Hadronic object reconstruction & classification is one of the most promising settings for cutting-edge machine learning and artificial intelligence algorithms at the LHC. In this contribution, highlights of ML/AI applications by ATLAS to QCD and boosted-object identification, MET reconstruction and other tasks will be presented.
Speaker: Maria Hernandez Sanz -
17
Hunting for H+→τν with the ATLAS Run 2 data
Charged Higgs bosons are predicted in many extensions to the Standard Model and represent a direct signature of an extended Higgs sector. This talk presents the ATLAS Run 2 results of searches for charged Higgs bosons decaying to a tau lepton and a neutrino, using 140fb-1 of proton-proton collision data recorded at √s = 13 TeV. Depending on the charged Higgs boson mass, it is produced either in top-quark decays or in association with a top-quark. During this process, a W boson is also produced, which decays either hadronically or semi- leptonically. Depending on this decay, the search targets τ+jets or τ+lepton final states. The tau lepton decays into a neutrino and hadrons are considered. Discrimination between signal and background is based on neural network parametrized as a function of charged Higgs mass. No significant excess above the Standard Model prediction is observed and exclusion limits are set on the charged Higgs production cross-section times branching ratio at 95% confidence level for the whole examined mass range.
Speaker: Julia Leszczynska (Polish Academy of Sciences (PL)) -
18
Global Particle Transformer for boosted jet tagging and mass regression in CMS
We introduce the Global Particle Transformer (GloParT), a large-scale pretrained model developed within the CMS experiment for boosted-jet tasks. Initiated in mid-2022, GloParT is designed as a “foundation model” to support a wide range of tagging and regression applications. It is pretrained on jets from over 300 categories, learning a rich and generalizable representation of jet substructure. We demonstrate effective fine-tuning of GloParT for multiple downstream tasks, including tagging Standard Model top, W, and Z jets, as well as a “scouting GloParT” variant adapted to the scouting data stream for analysis and HLT applications. The model and its fine-tuned variants are now deployed within CMS. GloParT outperforms previous approaches across a range of simulation benchmarks and has already been used in several physics analyses, including the recent public HH→4b result. We conclude by discussing the broader potential of universal pretraining in high-energy physics and directions for future development.
Speaker: Congqiao Li (Peking University (CN)) -
19
Heavy-flavor production inside jets with double-muon tagging in pp collisions at √s = 5.02 TeV
Heavy-quark production inside jets provides a sensitive probe of parton shower evolution and flavor creation in QCD. In particular, constraining $g \to c\bar{c}$ and $g \to b\bar{b}$ splittings in pp collisions is important both for understanding heavy-flavor production in vacuum and for establishing a reference for future studies in heavy ion collisions, where long-distance gluon splitting into heavy-quark pairs is predicted to be enhanced by in-medium parton interactions. Muon-based tagging offers a clean, high-purity way to isolate double-heavy-flavor topologies inside jets, with particular sensitivity to collinear gluon-splitting configurations, and is well suited to both the pp environment and future high-multiplicity heavy-ion collisions.
In this contribution, CMS presents a study of double-muon-tagged jets in pp collisions at $\sqrt{s}=5.02$ TeV. The measurement probes dimuon observables sensitive to the kinematic structure of heavy-flavor production within jets. The selected sample is dominated by double-$b$ production, enabling a characterization of heavy-flavor jet topologies associated with gluon splitting. These results provide a new benchmark for heavy-flavor production in jets in pp collisions and establish a promising path toward future studies of heavy-quark production and its modification in nuclear collisions.
Speaker: Dr Gian Michele Innocenti (Massachusetts Inst. of Technology (US)) -
20
Evidence of the dead-cone effect in bottom quark-initiated jets at the CMS experiment
The suppression of collinear gluon emissions from a massive quark, or the “dead- cone effect,” is the primary manifestation of quark mass effects in parton showers. A differential measurement is presented of the emission density of bottom quarks, studied as a function of the angular separation between the emissions from the same emitter. The measurement is performed in bottom quark-initiated jets with transverse momenta between 40 and 200 GeV originating from top quark and top antiquark production. A proton-proton collision data set recorded by the CMS experiment at √s = 13 TeV with an integrated luminosity of 59.8 fb$^{−1}$ is analyzed. The measurement of the emission density of bottom quarks strongly favors parton shower models that include the dead-cone effect.
Speaker: Jennifer Roloff (Brown University (US)) -
21
Search for new particles decaying into top quark-antiquark pairs in proton-proton collisions at √s = 13 TeV
A search for new particles decaying to top quark-antiquark pairs is performed using proton-proton collision data at a centre-of-mass energy of 13 TeV. The data set recorded with the CMS detector between 2016 and 2018 is used, corresponding to an integrated luminosity of 138 fb − 1 −1 . Final states with 0, 1, and 2 leptons are analyzed, covering all decay modes of the top quark-antiquark pairs. Heavy Z' bosons with relative widths of 1, 10, and 30% are excluded for masses in the ranges 0.4 − −4.8, 0.4 − −6.2, and 0.4 − −7.4 TeV, respectively. A Kaluza − −Klein gluon in the Randall − −Sundrum model and a dark-matter mediator are excluded for masses between 0.5 − −5.5 and 1.0 − −4.2 TeV, respectively. These results set the most stringent limits to date for the considered models in the t t ˉ t t ˉ final state. In addition, in the two-Higgs-doublet models, upper limits are set on the coupling strength modifier for scalar and pseudoscalar Higgs bosons with relative widths of 2.5, 10, and 25% in the mass range of 0.5 − −1.0 TeV.
Speaker: Haifa Rejeb Sfar (The State University of New York SUNY (US)) -
22
Searches for Heavy Vector-Like Quarks with ML-based jet-taggers in CMS
Vector-Like Quarks (VLQs) are a central feature of many BSM scenarios, including Composite Higgs and Extra-Dimensional models. These heavy fermions offer a natural solution to the hierarchy problem and provide a framework for understanding the Standard Model flavor structure. Unlike chiral fourth-generation quarks, VLQs are consistent with current Higgs boson measurements as their masses are not derived from Yukawa couplings. At the LHC, VLQs can be produced in pairs via the strong interaction or singly through electroweak processes. This talk reviews recent VLQ searches performed by the CMS Collaboration, with a focus on final states containing highly boosted objects. We discuss the application of state-of-the-art machine learning algorithms designed to identify the collimated hadronic decay products of W, Z, H bosons, and top quarks. These advanced jet-tagging techniques significantly improve the sensitivity to high-mass VLQs, and the latest exclusion limits from CMS are presented
Speaker: Petar Maksimovic (Johns Hopkins University (US)) -
23
Search for low-mass resonances decaying into a quark-antiquark pair produced with an initial state photon at √s = 13 TeV
A search for low-mass resonances decaying into a quark-antiquark pair in the mass range of 10 to 150 GeV is presented. This search uses data from the LHC in proton-proton collisions at √s = 13 TeV, collected by the CMS detector at the CERN LHC in 2016–2018, corresponding to an integrated luminosity of 138 fb−1. The analysis strategy makes use of an initial state photon recoiling against the resonance. As a result the resonance is produced with high transverse momentum, and the quark-antiquark decay products are reconstructed as a single large-radius jet with an internal two-pronged structure. A machine learning algorithm is used to distinguish such jets from single-prong jets. The results are interpreted in the context of a theoretical spin-1 vector boson, a leptophobic Z′, which is a dark-matter mediator candidate. No significant excess is observed when searching for this signature above the Standard Model background in the jet mass spectrum. Upper limits at the 95% confidence level are set on the coupling strength of the Z′ decaying to quark-antiquark pairs. This represents the most sensitive search of low mass resonances to quark-antiquark pairs to date in this mass range.
Speaker: Adam Albert Kobert (Rutgers State Univ. of New Jersey (US)) -
24
Pushing the Discovery Frontier: Extended Scalar Sector Searches with ML-based Jet Taggers at CMS
The discovery of the 125 GeV Higgs boson opened a new era in the search for heavy resonances at the LHC. Many BSM scenarios, such as 2HDM, 2HDM+S, TRSM, and NMSSM, predict extended scalar sectors with additional Higgs-like states. These models often produce final states with high-pT Higgs bosons, requiring advanced jet-tagging techniques. This talk presents the latest searches from the CMS Collaboration for a heavy scalar (X) decaying into a SM-like Higgs boson (H) and another scalar, which could be a BSM scalar (Y) or another SM-like Higgs boson. We highlight the critical role of state-of-the-art ML-based jet-taggers in identifying boosted decay topologies, which significantly enhance the sensitivity over traditional methods. A summary of recent results is provided, showcasing the most stringent constraints to date on several extended scalar benchmarks.
Speaker: Christos Leonidou (University of Cyprus (CY)) -
25
Search for nonresonant triple Higgs boson production in the six b quark final state in proton-proton collisions at CMS
A search for nonresonant triple Higgs boson production in the six b quark final state is performed using proton-proton collisions at √s = 13 TeV corresponding to an integrated luminosity of 138 fb−1 recorded by the CMS experiment. Each Higgs boson is reconstructed either from two small-radius jets (resolved) or a single large-radius jet (merged). No significant excess of events over the Standard Model background prediction is observed. Observed (expected) 95% confidence level upper limits on the signal cross section relative to the Standard Model expectation are set at 588 (572), corresponding to a signal cross section upper limit of 44 (43) fb. Assuming the quartic coupling modifier κ4 = 1, the observed (expected) constraint on the trilinear coupling modifier κ3 is −7.4 < κ3 < 12.4 (−6.4 < κ3 < 11.2) and, assuming κ3 = 1, the observed (expected) constraint on the quartic coupling modifier is −177 < κ4 < 185 (−177 < κ4 < 183). These are the most stringent constraints to date on nonresonant triple Higgs boson production and its quartic self-coupling, and provide the first probe, in a direct nonresonant triple Higgs boson search, of the perturbative unitarity boundary in the (κ3, κ4) plane.
Speaker: Xinyue Geng (Peking University (CN)) -
26
Measurement of the differential cross section with respect to jet mass in Z+jets eventsSpeaker: Aritra Mandal (The State University of New York SUNY (US))
-
14
-
-
-
Boosted-object tagging
-
27
Transforming Boosted Jet Tagging with the ATLAS Detector: ML algorithms and their performance
The identification of Higgs bosons with high transverse momenta is essential for the Higgs-boson research programme at the LHC, reaching across cross-section measurements in high transverse momentum regimes, Higgs-boson pair production searches, and searches for new heavy resonances. We present recent advances in Higgs-boson identification algorithms for cases where the Higgs-boson decay products are captured in a single large-radius jet using the ATLAS detector. These advances are driven by state-of-the-art machine learning techniques based on transformer architectures, providing up to a factor of 3 improvement compared to sub-jet tagging based approaches. The performance of these large-radius taggers for different Higgs-boson decay modes will be presented, discussing the rejection of large-radius jets that originated from top-quark decays, gluon splitting and W-boson decays. A brief outlook on the calibration strategies will be given as well.
Speaker: Osama Karkout (Nikhef National institute for subatomic physics (NL)) -
28
Boosted jet tagging and calibration in the CMS experiment
This talk summarizes recent progress in boosted object tagging in the CMS experiment, focusing on developments targeting Run 3 analyses. Updates to deep learning–based taggers are presented, including improvements in training strategies, input representations, and robustness to pileup and detector effects. Advances in calibration and systematic uncertainty evaluation are discussed, with emphasis on high-pT regimes. A dedicated focus is given to boosted tau identification, including adaptations of DeepTau-like approaches to collimated decay topologies. Their performance is evaluated in terms of efficiency and background rejection, and compared to standard resolved tau reconstruction.
Speaker: Congqiao Li (Peking University (CN)) -
29
The study of boosted jet identification and tagging at CMSSpeaker: Donato Troiano (Universita e INFN, Bari (IT))
-
30
Lund Plane to Bloch (LP2B) Encoding for Object and Polarization Tagging with Quantum Jet Substructure
The application of quantum algorithms to jet substructure analysis is of growing interest as Noisy Intermediate-Scale Quantum (NISQ) hardware continues to mature in qubit count and gate depth. Jet substructure remains essential for addressing challenges at the LHC and beyond, notably object classification and polarization tagging. However, existing quantum machine learning approaches typically rely on data representations that suffer from infrared and collinear (IRC) unsafety, sensitivity to non-perturbative effects, or poor scalability.
In this talk, we introduce the Lund Plane to Bloch (LP2B) [1] encoding, which maps a theoretically clean and robust representation of jet kinematics directly into qubit states. Leveraging this encoding, we implement a Quantum Tree-Topology Network (QTTN) that natively embeds the hierarchical structure of the Lund tree. We evaluate the QTTN across multiple benchmarks, comparing it with classical machine learning architectures and the standard "one particle - one qubit" (1P1Q) encoding on polarization, W boson, and top quark tagging tasks, including in the low-data regime. The results show that, despite its low parameter count, the QTTN achieves competitive performance with classical baselines, demonstrates enhanced sensitivity compared to the 1P1Q encoding, improves the performance-to-cost trade-off, and exhibits enhanced stability in low-data regime and reduced sensitivity to generator-specific parton shower and hadronization models. Finally, the QTTN is validated on real quantum hardware using a 3-qubit solid-state NMR SpinQ device.[1] https://arxiv.org/abs/2604.18613
Speaker: Tommaso Tedeschi (Universita e INFN, Perugia (IT))
-
27
-
10:20
Coffee Break
-
Large-radius jet calibration and regression
-
31
Calibration and uncertainties of large-radius jets with ATLAS LHC Run2 data
Large radius jets are useful objects to describe massive resonances produced with a high boost which decay fully hadronically. During the Run~2 run of the LHC, the ATLAS experiment developed a new algorithm to define large-radius jets, clustering Unified Flow Objects (UFO) constituents and applying the constituent subtraction (CS), soft killer (SK) and soft drop (SD) techniques. In this paper, the calibration of the jet energy and mass of the ATLAS large-radius jets is described. The calibration procedure employs both simulation-based and data-based (in-situ) methods and validates the jet detector response in data using the full Run~2 dataset. The residual uncertainties on the jet energy and mass scale are close to 1% for transverse momenta up to 1 TeV and are controlled to 2-3\% up to 2 TeV , enabling sensitive searches and precise studies of boosted object production in the Standard Model.
Speaker: Yicong Huang (Charles University (CZ)) -
32
Transformer models for kinematic regression of R=0.4 and R=1.0 bottom-quark jets in ATLAS
Decays of the Higgs and Z bosons to bottom quarks are essential for measurements of the bottom quark Yukawa coupling, the Higgs self-coupling, and production modes in highly boosted phase spaces. In ATLAS, these decay products are reconstructed as jets using a clustering radius parameter of R=0.4 and R=1.0 for resolved and boosted decays, respectively. Improvements in flavour tagging, which labels jets as originating from the production of bottom-quark hadrons, have dramatically improved the sensitivity to H->bb and Z->bb decays in recent years by suppressing backgrounds originating from light jets. This talk will summarize the next step in improving analysis sensitivity by applying the transformer architecture to regress the transverse momenta and mass of small- and large-radius jets based on tracks, calorimeter clusters, and leptons within jets. The increase in resonant decay invariant mass resolution is improved up to 30% with respect to the standard jet calibration. The regression models are calibrated in-situ using the method of direct momentum balance in events with jets recoiling against Z->ll or a photon for R=0.4 jets and events targetting Z->bb recoiling against a photon or multiple jets for R=1.0 jets. The jet energy and mass scales for R=1.0 jets in simulation is compared to data. These models therefore offer significant enhancement of resonant X->bb signals that can be directly translated to improvements in physics measurement and search sensitivity.
Speaker: Snigdho Chakraborty (University of Warwick (GB))
-
31
-
Measurements and searches with boosted objects
-
33
Boosted W/top mass and substructure at ATLAS and CMS
This talk presents a comprehensive overview of recent measurements of the mass and substructure of boosted top quarks and W bosons using full Run 2 data from ATLAS and CMS. Particular emphasis is placed on the extraction of the top quark mass in highly collimated topologies, contrasting precision determinations from ATLAS with advanced reclustering strategies used by CMS, such as the use of XCone to enhance reconstruction quality. The presentation also summarizes recent progress toward precision measurements of the W boson mass in all-hadronic final states, where the soft-drop algorithm is applied to mitigate pileup and radiation effects. Beyond mass extractions, it highlights detailed evaluations of jet substructure, including simultaneous measurements of N-subjettiness observables across light-flavour quark, gluon, W boson, and top quark jets. These studies are complemented by in-depth analyses of top jet substructure and the latest characterizations of hadronic emissions using the Lund jet plane. Collectively, these results showcase the ongoing refinement of algorithmic techniques and observables that are essential for advancing precision SM measurements at the LHC.
Speaker: Conrado Munoz Diaz (Riga Technical University (LV)) -
34
Extended Scalar Sector and Heavy Vector-Like Searches with ML-based jet-taggers in CMSSpeaker: Christos Leonidou (University of Cyprus (CY))
-
35
Search for boosted Higgs boson productions at ATLAS
Measurements of Higgs boson productions in the very high pT regime is of great interest because it is sensitive to potential new physics beyond the Standard Model. Such measurements are typically carried out in channels with large branching ratios (such as H->bb) due to the low rate. The decay products of the Higgs boson will be so collimated in the high pT regime that they can be reconstructed as a large-radius jet. Reconstruction, identification and calibration of large-radius Higgs boson candidates is an important experimental challenge. In this presentation, we will present the latest ATLAS searches for boosted Higgs boson productions in various production modes and decay channels, featuring the boosted jet performance from the ATLAS experiment.
Speaker: Gabriel Facini (University of London (GB))
-
33
-
12:30
Lunch Break
-
Flavour tagging
-
36
Transforming Flavour Tagging with the ATLAS Detector: ML tools and calibration techniques
The identification of jets containing b-hadrons is essential for many physics analyses at the LHC, including precision measurements of Higgs boson and top-quark processes, as well as searches for physics beyond the Standard Model. We present recent improvements in the discrimination of b-jets from jets originating from lighter quarks using the ATLAS detector. These advances are driven by state-of-the-art machine learning techniques based on transformer architectures. Their performance is well modelled by the ATLAS simulation, as demonstrated through dedicated calibration studies, the results of which will be presented. Compared to previous algorithms, the transformer-based approach improves the rejection of c-jets (light-jets) by factors of 3.5 (1.8) at a b-jet tagging efficiency of 70%. We also discuss the latest version of this algorithm and its expected performance at the High-Luminosity LHC (HL-LHC).
Speaker: Greta Brianti (Nikhef) -
37
Jet flavor tagging developments and performance in the CMS experiment
Innovation in jet tagging techniques to identify (heavy) quark flavours or gluons, as well as hadronic tau reconstruction and identification, has been an important driver to maximally exploit the physics potential of LHC data, and remains an active area of study in CMS. In this talk we present the latest algorithm developments and performance results.
Speaker: Philipp Gadow (Hamburg University (DE)) -
38
Jet flavour tagging studies for e+e- collisions at FCC-ee
Jet flavour identification algorithms are essential to maximise the physics output in e+e- collisions at the Future Circular Collider (FCC-ee). Within the broad FCC-ee physics programme, jet flavour tagging plays a central role in QCD studies as well as in precision measurements of the Higgs boson, electroweak bosons, and top quark, owing to the dominance of hadronic decays of the heaviest Standard Model (SM) particles. Highly efficient discrimination among bottom-, charm-, strange-, and gluon-initiated jets enables precisely measuring decay channels that are inaccessible at the LHC, thereby significantly enhancing the sensitivity of precision SM measurements and searches for new physics at FCC-ee. In this contribution, we present recent developments in jet flavour tagging based on machine learning techniques applied to representative FCC-ee physics studies. In addition to excellent performance in b- and c-jet identification, future taggers are expected to achieve high discrimination for strange-quark jets, opening the possibility of measuring the strange-quark Yukawa coupling as well as probing rare flavour-changing neutral current processes such as $Z \to bs$. Potential strategies for calibrating jet tagging performance will be also discussed.
Speaker: Apranik Fatehi (University of Hamburg - CERN - DESY) -
39
Hidden Flavor Information in Kinematics of Particle Flows at the Large Hadron Collider
Modern jet taggers operating directly on jet constituents have shown impressive performance, yet the high-level physics features they exploit to outperform classical methods remain only partially understood. In this paper, we discuss hidden particle flavor information imprinted in the kinematics of jet constituents; this flavor information is less accounted for in constructing high-level features that rely solely on the constituent kinematics. We argue that the flavor information and constituent kinematics are not independent because of angular resolution heterogeneity and particle-dependent response of detector subsystems, and hence, the modern jet taggers can exploit this flavor information even when the constituent-type information is masked out. To demonstrate accessibility of the flavor information, we first compare two Particle Transformers classifying flavor-specific top jets in a constituent type-blinded setup, i.e., $t \rightarrow bu\bar{d}$ jets vs.~QCD jets and $t \rightarrow bc\bar{s}$ jets vs.~QCD jets. The flavor asymmetry in the classification performance is a sign that Particle Transformer has access to the hidden flavor information. As the particle flow object type is the key detector-level quantity that carries the flavor information, we then explicitly show that object types can be inferred from the kinematics of neighboring constituents in order to demonstrate a correlation between flavor and kinematics. Our findings provide insight into why modern jet tagging architectures perform well, and highlight the importance of understanding the interplay between particles and detector effects when interpreting highly capable neural networks for jet classification.
Speaker: Sung Hak Lim (IBS CTPU-PTC) -
40
Calibration of a Transformer-Based Quark/Gluon Tagger in ATLAS
Distinguishing quark-initiated jets from gluon-initiated jets is impactful for numerous LHC physics analyses, including precision Higgs boson measurements and searches for a production of two Higgs bosons in vector-boson fusion topologies. This talk presents DeParT, a transformer-based quark/gluon tagger developed by ATLAS that exploits the full information from jet constituents reconstructed in both the tracker and calorimeter systems, operating across an extended kinematic phase space. The efficiency measurement in data and tagger calibration is performed using dijet events from the LHC Run 2 and Run 3 proton--proton collision data. A key advancement is the deployment of the jet topics method, a data-driven technique that reduces reliance on Monte Carlo modelling compared to traditional approaches, yielding up to 20% smaller systematic uncertainties in some kinematic regions. The resulting efficiency scale factors enable robust application of the tagger across the ATLAS physics program.
Speaker: Samuel Jankovych (Nikhef National institute for subatomic physics (NL))
-
36
-
15:40
Coffee Break
-
Heavy flavour and new physics in jets
-
41
The study of heavy flavour jets at CMSSpeaker: Dr Gian Michele Innocenti (Massachusetts Inst. of Technology (US))
-
42
Inclusive Dijet and Heavy Flavour Jet Production at ATLAS
We present new jet production results from ATLAS. Measurements of multi-differential dijet production using the full Run 2 data set are compared to state-of-the-art predictions. In addition a new measurement of W production in association with b-jets provides an important test of perturbative quantum chromodynamics (pQCD) in the presence of heavy quarks.
Speaker: Lee Sawyer (Louisiana Tech University (US)) -
43
Phenomenology of heavy flavour jet angularities at hadron colliders
We compute resummed and matched predictions for jet angularities in hadronic $Z+$jet events, where the jet is initiated by a $b$-quark. The analysis is performed both with and without grooming the candidate jets using the SoftDrop algorithm. Mass effects are consistently included at both fixed-order and resummed levels.
Our theoretical predictions also incorporate non-perturbative corrections from the underlying event and hadronization, implemented through parton-to-hadron transfer matrices extracted from dedicated Monte Carlo simulations with Sherpa. Finally, we compare our results with previous implementations in order to quantify the impact of mass effects.Speaker: Andrea Ghira (TUM) -
44
Lund jet planes of hidden and double-open heavy-flavor jets
Measurements of quarkonium-in-jet fragmentation have established that hidden heavy-flavor hadrons, such as $J/\psi$ ($c\bar c$) and $\Upsilon$ ($b\bar b$) states, are produced copiously within active jet showers. This has motivated new developments in parton shower algorithms, opening a broader avenue for heavy-flavor jet substructure studies. We present a comparative phenomenological study of quarkonium-tagged and double-open-heavy-flavor-tagged jets using the Lund jet plane. Events are simulated with Pythia8 for proton-proton collisions at LHC energies, using the recent OniaShowers module to generate quarkonium-tagged jets. The Lund jet plane resolves successive stages of the jet shower associated with heavy-flavor gluon splitting, from the gluon-rich cascade preceding the $g\to Q\bar Q$ splitting, to the splitting itself, the subsequent radiation of the heavy-quark antenna, and ultimately hadronization. Collinear emissions along the quarkonium-following prong are found to be strongly suppressed relative to the corresponding double-open-heavy-flavor case, reflecting the formation of color-singlet heavy-flavor bound states. Comparisons to a gluon-jet baseline are used to visualize mass effects in the collinear region and to provide a reference Casimir color factor in the hard, wide-angle region of the plane. $J/\psi$-tagged jets exhibit Lund plane densities compatible with the gluon-jet baseline for hard emissions, suggesting a possible path to test the radiative behavior of intermediate color-octet states in general. Overall, these studies show how quarkonium-tagged and double-open-heavy-flavor jets can jointly provide new insight into the overall radiation pattern associated to jet showers.
Speakers: Cristian Baldenegro (Massachusetts Inst. of Technology (US)), Florian Damas (UNESP - Universidade Estadual Paulista (BR)), Matthew Nguyen (Centre National de la Recherche Scientifique (FR)), Paul Ghinderelli -
45
Jet Substructure Observables as Probes of Heavy Neutrino Polarization in Boosted W′ Decays at the HL-LHC
We present a jet substructure-based polarimetry framework for boosted heavy neutrino fat jets at the HL-LHC, with the goal of probing the chiral structure of heavy $W'$ interactions. In the boosted regime, a heavy neutrino $N$ produced via $pp \to W' \to \ell N$ and decaying as $N \to \ell jj$ gives rise to a collimated final state. This final state can be reconstructed as a large-radius 3-pronged fat jet containing an embedded lepton. This topology provides a unique opportunity to access polarization information directly from jet substructure.
The central contribution of this work is the construction of three novel polarization-sensitive jet substructure observables namely, $z_{\ell}$, $z_{\theta}$, and $z_{k}$. These variables are physically motivated by the spin-analyzing power of the decay products of $N$ and are designed to encode the underlying chiral structure of the interaction, distinguishing between left-chiral $(V-A)$ and right-chiral $(V+A)$ couplings in a model-independent manner. When combined into a multivariate discriminant using boosted decision trees and evaluated with a $\mathrm{CL}_{s}$-type profile likelihood estimator, these observables achieve a discrimination significance of $1.6\sigma$-$2.8\sigma$ between the two chiral coupling hypotheses at an integrated luminosity of $3~\mathrm{ab}^{-1}$.Speaker: Songshaptak De (Jožef Stefan Institute)
-
41
-
-
-
New measurements from RHIC and archived data
-
46
Jet Results in p+p Collisions at STAR
Jet production in high-energy collisions at RHIC provides direct access to the spin and flavor structure of the nucleon, as well as to the mechanisms of spin-dependent fragmentation and hadronization. This talk will present recent STAR results on jet production in proton-proton collisions, with emphasis on cold QCD and spin physics. The topics to be discussed include jet spin asymmetries, hadron-in-jet measurements, spin transfer to $\Lambda$ and $\bar{\Lambda}$ hyperons, and transverse polarization of $\Lambda$ and $\bar{\Lambda}$ hyperons inside jets. They provide valuable constrains to transverse-momentum-dependent (TMD) parton distribution functions and fragmentation functions. In particular, the polarizing fragmentation function, which describes the production of transversely polarized hyperons from unpolarized partons during hadronization, can be directly constrained by the hyperon polarization in jets. These measurements connect perturbative QCD, nucleon spin structure, and spin-related hadronization phenomena. The talk will highlight the role of STAR jet measurements in bridging these areas and discuss their relevance for future studies at the Electron-Ion Collider.
Speaker: Taoya Gao (Shandong University) -
47
Recent jet and event-shape observable results from the Electron-Positron Alliance
Electron-positron collisions provide a uniquely clean environment for precision studies of jets and Quantum Chromodynamics, free from hadronic initial-state effects and with a well-defined hard-process energy scale. Recent progress in archival-data preservation and modern analysis techniques has opened a new program of jet and energy-flow measurements using LEP data, allowing contemporary jet-substructure tools to be applied to high-resolution $e^+e^-$ events. In this talk, we present recent results from the Electron-Positron Alliance based on archived ALEPH and DELPHI data. The measurements include anti-$k_T$ jet spectra across different resolution parameters, substructure observables, energy-energy correlators, and thrust in hadronic $Z$ decays and related measurements. These observables probe perturbative radiation, hadronization, and the transition from partons to hadrons in a single clean experimental setting. The results are compared with modern event generators and, where available, perturbative and resummed QCD calculations. Together, these measurements establish LEP archival data as a precision laboratory for modern jet physics. They provide stringent constraints on event-generator modeling, new baselines for jet-substructure measurements at hadron colliders, and important inputs for future $e^+e^-$ collider programs. They also demonstrate the broader potential of the "recycling frontier": extracting new QCD information from preserved collider datasets using modern theoretical and experimental tools.
Speaker: Shuangyi (Bill) Zhou (Vanderbilt University) -
48
Measurement of Spin and Quantum Correlations in Z → τ+ τ- Decays at LEP with DELPHI
Precision studies of $τ⁺τ⁻$ production at the Z pole provide a clean environment for investigating electroweak spin correlations and quantum information observables. Using archived LEP-1 data collected by the DELPHI experiment, the process $e⁺e⁻ → Z → τ⁺τ⁻$ is well measured, but the presence of multiple neutrinos in $τ$ decays limits reconstruction of the $τ$-pair rest frame. This constrains the precision of spin-dependent measurements.
We present a machine-learning reconstruction method based on an event-level foundation model that predicts neutrino momenta using detector-level observables and kinematic constraints. The reconstructed $τ⁺τ⁻$ rest frame improves the resolution of spin-sensitive observables and enables a measurement of the $Z → τ⁺τ⁻$ spin density matrix. These results provide a basis for quantum correlation studies using archived $e⁺e⁻$ collider data and demonstrate the applicability of machine-learning–based multi-neutrino reconstruction in precision electroweak analyses.
Speaker: Chen-Hua Hsu (National Taiwan University (TW)) -
49
Extraction of Quark and Gluon Jets with CMS Open Data
We present a measurement of a set of generalized jet angularities using proton-proton collision data at $\sqrt{s} = 7$ and $13$ TeV from the CMS Open Data. The measured distributions are unfolded to particle level to correct for detector resolution and acceptance effects, accompanied by evaluations of experimental and methodological systematic uncertainties. Measurements at two or more energies can then be combined to yield distributions of any jet property separated into quark and gluon jet samples based on a previous Monte Carlo study.
Speaker: Van Dung Le (Helsinki Institute of Physics)
-
46
-
10:20
Coffee Break
-
Jet quenching and heavy-ion substructure
-
50
Jet substructure measurements in p+p and heavy-ion collisions at STAR
Jets are multiscale objects that provide a direct connection between partons and hadrons, making jet substructure measurements a powerful probe of both perturbative and non-perturbative QCD dynamics. At STAR, a broad set of jet substructure observables, including SoftDrop-groomed observables and N-point Energy Correlators (ENCs), is used to investigate parton shower evolution and hadronization mechanisms. These observables maintain a close connection to the underlying partonic dynamics, enabling detailed comparisons with first-principles theoretical calculations. Furthermore, charge-sensitive observables, such as charge-weighted ENCs, provide additional sensitivity to the hadronization process.
In this talk, we present recent STAR measurements of SoftDrop observables and ENCs for different jet transverse momenta and jet radii in both $p$+$p$ and heavy-ion collisions (Au+Au, Zr+Zr, Ru+Ru). These measurements provide new insights into the structure of jets and their modification in the QCD medium created at RHIC.
Speaker: Monika Robotková (Nuclear Physics Institute, Czech Academy of Sciences) - 51
-
52
First unfolded measurement of the jet quenching effect with Z+jet events in PbPb and pp collisions at 5.36 TeV with Run 3 data
The production of jets in association with Z bosons offers a uniquely clean probe of the quark-gluon plasma (QGP), as the Z boson's decay products bypass strong medium interactions, preserving the initial hard-scattering kinematics. We present the latest measurement of Z+jet correlations in the dimuon channel using the new 5.36 TeV Run 3 data collected by the CMS experiment in 2023 and 2024. The study focuses on the momentum imbalance between the Z boson and the recoiling jet, providing a direct and sensitive look at medium-induced energy loss and the in-medium evolution of partons. The unfolded $x_{Zj}$ distribution, compared with state-of-the-art theoretical calculations, offers crucial tests of the microscopic structure of the QGP. By leveraging the high statistics and increased energy of Run 3, this measurement provides new experimental leverage to constrain theoretical models of parton energy loss and the transport properties of the QGP. Furthermore, the unprecedented statistical precision of the anticipated full Run 3 dataset will pave the way for detailed studies of jet substructure in this clean Z+jet channel.
Speaker: Raffaele Delli Gatti (Universita e INFN Trieste (IT)) -
53
Probing jet quenching via jet substructure measurements in Pb+Pb collisions with ATLAS
In ultra-relativistic heavy-ion collisions, jet yields are suppressed relative to proton–proton baselines due to partonic energy loss in the quark–gluon plasma. This phenomenon, known as jet quenching, reflects the interaction of hard-scattered partons with the hot and dense medium created in the early stages of the collision. This ATLAS contribution presents a set of measurements that go beyond inclusive observables by probing the connection between jet quenching and jet fragmentation using substructure techniques. The results include measurements of large-radius jet substructure using track and sub-jets, studies of jets recoiling against isolated photons, and measurements of b-jet suppression, providing sensitivity to the flavour dependence of partonic energy loss. Jet quenching is quantified via the nuclear modification factor, $R_{AA}$, measured as a function of centrality, jet transverse momentum, and a range of substructure observables. Together, these measurements offer complementary constraints on quark- and gluon-initiated jets and their modification in the medium.
Speaker: Petr Balek (AGH University of Krakow (PL)) -
54
Bottom-up approach to describe groomed jet data in heavy-ion collisions
The theoretical interpretation of jet observables in heavy-ion collisions is a complex task due to the intricate interplay of perturbative and non-perturbative effects. One way to reduce this complexity is to groom away soft, wide-angle radiation so that perturbative dynamics dominates. Even in this simplified scenario, there are competing explanations for the physical origin of the measured medium-induced modifications. In this paper, we present a minimal approach to compute groomed substructure observables. The core idea is to treat medium effects as an effective energy shift of the hard, vacuum-like substructure. This energy shift includes a gradual onset of colour decoherence effects and thus depends on the jet substructure itself. We first study a NLO-exact dijet configuration in vacuum and apply radiative energy-loss to the two subjets. We find that this minimal setup already captures the narrowing trend of groomed observables but it's not able to quantitatively describe the existing data. Next, we match the NLO matrix-element to a leading-logarithm accurate parton shower and perform a clustering algorithm to recover a two-prong system to which we again apply the energy-loss distribution. Despite its simplicity, the model results in a very good theory-to-data agreement (within 10%) for a broad range of observables including both ALICE and ATLAS kinematics. We also examine the discriminating power of groomed jet data in terms of colour decoherence effects and find that substructure-dependent energy loss yields an overall better agreement.
Speaker: Diogo Costa (Universidad de Granada)
-
50
-
12:30
Lunch Break
-
Excursion and conference dinner
-
-
-
Event shapes, showers and substructure
-
55
Event Isotropy in perturbative QCD
It has recently been proposed that collider events can be equipped with a metric, the Energy Mover’s Distance (EMD), which allows one to rephrase multiple aspects of jet physics in a geometric language. Further, the EMD can be exploited to define new observables that measure the distance between a given event and an idealised one. For instance, event isotropy quantifies the resemblance of an event to a uniform energy distribution. We present the first field-theoretical description of the event isotropy distribution in electron-positron collisions. Our calculation includes the all-order resummation of soft and collinear logarithmic contributions at next-to-leading log accuracy, matched to next-to-leading order corrections at fixed order.
Speakers: Daniele Atzori (LPTHE), Daniele Atzori (LPTHE Paris & INFN Genova) -
56
Jet fragmentation at next-to-single log (NSL) for general cuts
It has been shown recently that defining a jet by a transverse-momentum style resolution parameter produces a fragmentation function that evolves with the standard timelike $\overline{\rm MS}$ DGLAP kernels, to next-to—single logarithmic ($\alpha_s^n L^{n-1}$) accuracy. In contrast the NSL fragmentation function for a purely angular cut like a small jet radius deviates from time-like DGLAP evolution.
This raises the natural question of how the evolution depends on a general clustering cut which is neither pure transverse momentum nor angle. This is particularly relevant in the context of modern parton showers, which employ a variety of evolution variables and are aiming to reach general NNLL/NSL accuracy.
Here we obtain the closed-form NSL result — i.e. at the level of NLO DGLAP evolution — for a general cut. The result interpolates linearly between the time-like (standard) and space-like (angular) DGLAP limits, with the boost non-invariance of the cut as the interpolation parameter. It can be used as a potential test for future NNLL showers as well as being of direct phenomenological interest.
Speaker: Alexander Fraley (University of Manchester) -
57
Searching for New Physics Inside Jets with the Herwig 7 Generalised Parton ShowerSpeaker: Joon-Bin Lee (Seoul National University (KR))
-
58
Machine learning fully hadronic events with spectral functions
Fully hadronic events at hadron colliders are difficult to analyse because extra QCD radiation creates many possible jet combinations and varying jet multiplicities. I will discuss how the two-point correlation spectral function can provide a compact, permutation-invariant event representation for machine-learning analyses. As a benchmark, I apply this method to gluino pair production followed by decays to top quarks and neutralinos, and show that it can improve the expected gluino-mass reach compared with standard strategies.
Speaker: Kazuki Sakurai (University of Warsaw)
-
55
-
10:20
Coffee Break
-
Novel methods from information theory and ML
-
59
QCD Theory meets Information Theory
We begin by reviewing how NLL accuracy is achieved in modern parton showers, highlighting the recent Sherpa implementation, and then introduce our new information-theoretic matching framework to achieve beyond NLL accuracy. By minimizing a Kullback–Leibler functional under constraints set by precision QCD input observables (including theory uncertainties), we embed high-order predictions into fully differential, particle-level simulations with strictly positive event weights and the ability to impose multiple observable constraints simultaneously. We futher discuss HDSense a new efficient tool that approximates the full Fisher information matrix for observable selection to maximize sensitivity to underlying QCD parameters.
Speaker: Benoit Assi (Fermilab) -
60
When Do ML Uncertainties Fail? - A PDF fit case study
With the vast amount of data already produced by the LHC and the upcoming HL-LHC runs, reliable and precise uncertainty quantification for numerical tools, in particular neural networks, is essential. We present a systematic study of uncertainty estimation using various machine learning approaches, including neural networks, Gaussian processes, and neural tangent kernels.
As an underlying dataset, we use the T3 contribution to parton distribution functions (PDFs), based on datasets already employed in NNPDF, to provide a controllable yet well-understood framework for our analysis. We investigate how uncertainty estimates depend on the modelling strategy and its underlying assumptions, including the incorporation of physics knowledge. In addition, we test all methods in extrapolation regions and assess the trustworthiness of their uncertainty estimates.
For practical validation, we compare the T3 model predictions obtained with different numerical setups to deep-inelastic scattering data from HERA, providing a phenomenological benchmark for the predictions and their uncertainties. Our results highlight the strong method dependence of ML-based uncertainty estimates and underscore the need for careful validation when applying these techniques in collider phenomenology.Speaker: Nina Elmer (University of Cambridge) -
61
Using Agentic AI Without Losing the Physics: Lessons from LEP Data Analysis
Agentic AI systems are rapidly moving beyond information retrieval and code completion toward the execution of complex, multi-step technical tasks. For experimental particle physics, this creates both an opportunity and a challenge: AI agents can accelerate analysis implementation, validation, and documentation, but physics measurements require controlled reasoning, reproducibility, and expert judgment rather than merely producing results that appear plausible. In this talk, we discuss recent work by the Electron-Positron Alliance using archived LEP data as a controlled testing ground for agentic AI in collider analysis. As a proof of concept, AI agents were used under expert physicist direction to perform a measurement of the thrust distribution in archived ALEPH e+e- collisions at sqrt(s) = 91.2 GeV, including event selection, correction procedures, uncertainty studies, unfolding, and analysis-note writing. We also discuss exploratory applications to DELPHI data, including particle-identification and K/pi studies, where fewer pre-existing analysis blueprints are available. The central focus is NOT whether AI can generate analysis code quickly, but how it can be integrated into a physics-rigorous workflow. We emphasize a conservative operating model in which the AI agent accelerates technical implementation while the physicist retains ownership of analysis choices, validates each step, and controls the scope through small, inspectable tasks. These studies suggest that preserved e+e- datasets provide an ideal environment for developing practical standards for AI-assisted physics analysis: transparent workflows, human-in-the-loop validation, robust guardrails, and reproducible outputs. More broadly, they point toward a future theory-experiment loop in which AI agents assist with calculations, measurements, comparisons, and documentation, while scientific judgment remains firmly in human hands.
Speakers: Anthony Badea (University of Chicago (US)), Hannah Bossi (Massachusetts Inst. of Technology (US)) -
62
Going HyPER (and beyond): geometric machine learning methods for collider data analysis
Modern analysis of collider data relies heavily on machine learning. This talk will discuss geometric machine learning methods, focusing on the HyPER tool: applying hypergraph representation learning to reconstruct short-lived particles in collider data. The method is extended through generative ML techniques to include production processes featuring neutrinos, making the HyPER tool a comprehensive event reconstruction method. Studies on the impact of event reconstruction in signal-background classification will also be presented, as well as showing contrasting ML approaches to building event observables. The methodologies presented will be showcased in several different collider processes, and their application to real data discussed.
Speaker: Ethan Lewis Simpson (The University of Manchester (GB)) -
63
New event-wide strategies for increasing di-Higgs efficiency with Machine Learning at trigger level
The measurement of the trilinear Higgs boson self coupling, accessible through Higgs pair production, is one of the most important goals of the High-Luminosity LHC (HL-LHC), since it allows to probe the shape of the Higgs potential, a crucial prediction of the electroweak symmetry breaking mechanism. However, the low pair production cross-section is a challenge, namely for the online event processing and selection performed by the trigger system. Despite having the largest branching ratio, the fully hadronic HH->4b channel poses significant additional triggering difficulties, since the cross-section for b-quarks is approximately seven orders of magnitude larger than for the Higgs boson. Currently, most triggers employed at the High-Level Trigger (HLT) rely on b-tagging algorithms, which are both CPU expensive and need high thresholds to reduce background rate. This motivates the study of alternative signal selection techniques.
In this work, we aim to employ Energy Flow Polynomials (EFPs) computed at event-level and study their sensitivity to different event topologies, discriminating four-jet HH->4b signal events from dijet background, for an environment with pile-up of 200 proton-proton collisions per bunch crossing. The performance of a supervised linear classifier trained with the EFPs is benchmarked by two alternative techniques: a standard multijet trigger with a transverse momentum cut and a Deep-Sets-based neural network, implemented with Particle Flow Networks (PFN) . Deep Sets algorithms using hadronic jets and tracks as inputs for b-tagging have been employed by the ATLAS experiment in the HLT. In this work, we use only the four-momentum of the jets in the events as input to the Deep Sets. Jets in the simulated events were reconstructed using the anti-kt algorithm, with R=0.4, and their transverse momentum was smeared to reproduce the expected resolution of the HLT. We show that both event-level EFPs and PFN increase the signal selection efficiency by at least 30% for events below the selection threshold of a trigger based on a multijet transverse momentum cut, for the same background acceptance rate. Finally, we also show that, due to the strong correlation between the EFPs, event selection relies on a very limited number of variables per event, which is particularly relevant in the software trigger, where latency and computational resources are limited.
Speaker: Ana Rita Ferreira Carvalho (Instituto Superior Técnico and Laboratory of Instrumentation and Experimental Particle Physics (PT))
-
59
-
12:30
Lunch Break
-
Negative-weight mitigation for MC simulation
-
64
Resummed Distribution Functions: Making Perturbation Theory Positive and Normalized
Fixed-order perturbative calculations for differential cross sections can suffer from non-physical artifacts: they can be non-positive, non-normalizable, and non-finite, none of which occur in experimental measurements. We propose a framework, the Resummed Distribution Function (RDF), that, given a perturbative calculation for an observable to some finite order in $\alpha_s$, will ``resum'' the expression in a way that is guaranteed to match the original expression order-by-order and be positive, normalized, and finite. Moreover, our ansatz parameterizes all possible finite, positive, and normalized completions consistent with the original fixed-order expression, which can include N$^n$LL resummed expressions. The RDF also enables a more direct notion of perturbative uncertainties, as we can directly vary higher-order parameters and treat them as nuisance parameters. We demonstrate the power of the RDF ansatz by matching to thrust to $\mathcal{O}(\alpha_s^3)$ and extracting $\alpha_s$ with perturbative uncertainties by fitting the RDF to ALEPH data.
Speaker: Radha Mastandrea -
65
Cell Reweighting and Unbinned Validation for Negative-Weight Mitigation using Optimal Transport
As the accuracy of experimental results increases in high energy physics, so too must the precision of Monte Carlo simulations. Currently, event generation at next-to-leading order (NLO) accuracy in QCD and beyond results in the production of negatively-weighted events. The presence of these weights increases strain on computational resources by degrading the statistical power of MC samples, and can be pathological in the context of machine learning. We have developed a post hoc "cell reweighting" scheme that applies an IRC-safe metric in the multidimensional space of events so that nearby events are reweighted together, with the metric implemented using Optimal Transport techniques borrowed from computer vision to address this longstanding problem in computational particle physics. We compare the performance of the algorithm under different choices of metric and explicitly demonstrate its behaviour on simulated events with a Z boson and two jets produced at NLO accuracy.
To validate that these full phase-space reweightings preserve the physical fidelity of the underlying model — a task for which comparisons to marginalized 1D kinematic histograms can mask subtle biases — we additionally introduce an unbinned figure of merit based on the "Cross-Section-Mover's Distance," an Optimal Transport-based quantity that measures the work required to transform one theoretical prediction into another. This complementary metric provides a principled way to benchmark reweightings performed with different metric choices (e.g., Euclidean vs. Energy-Mover's Distance) and can be applied more broadly wherever phase-space reweighting biases must be studied in an unbinned way.
Speaker: Jennifer Roloff (Brown University (US)) -
66
Positive Smeared Matrix Elements at Next-to-Leading Order
The issue of negative weights in the simulation of particle collider events at higher orders in perturbation theory can significantly reduce numerical precision, for a given statistical sample size. Several methods for reducing or even eliminating negative event weights have been proposed, including resampling techniques that involve summing or ``smearing over'' nearby events on phase space to ensure positivity. Such methods have typically used machine learning algorithms to perform the resampling of the data ensemble, but effectively use no physics to inform it. We introduce an event smearing algorithm that exploits the universality of soft and collinear divergences in quantum chromodynamics, explicitly calculating all necessary components at next-to-leading order. We are able to show that the effect of this smearing does introduce numerical errors in the now-positive event weights, but the size of these effects are suppressed by powers of the smearing radius and are typically smaller than unknown next-to-next-to-leading order contributions. We demonstrate this procedure in two- and three-jet events in $e^+e^-$ collisions, and provide first results for its extension to the smearing of events at next-to-next-to-leading order.
Speaker: Rikab Gambhir (University of Cincinnati) -
67
The Latent Information Geometry of Jet Classification
Latent representations are an important theme in modern machine learning. Any network training with the notion of locality introduces a latent geometry which we can analyze with the help of differential geometry, specifically information geometry. We introduce the main concepts needed to analyze learned latent geometries, specifically curvature and nonmetricities, and show how they can be used for decoder and classifier geometries. We then apply our new methods to understand the physics behind binary quark-gluon classification.
Speaker: Sophia Vent
-
64
-
15:20
Coffee Break
-
Anomaly detection and model-agnostic searches
-
68
Recent Results from Anomaly Detection Searches on CMS
In the absence of direct evidence for new physics in targeted searches, model-independent strategies are becoming increasingly important. In this talk, we present recent results of model-agnostic searches that are facilitated by advanced machine learning techniques, opening a new avenue for unbiased detection of potential new physics signals.
Speaker: Petar Maksimovic (Johns Hopkins University (US)) -
69
Searches for new physics with the ATLAS detector
Many theories beyond the Standard Model (SM) have been proposed to address several of the SM shortcomings. Some of these beyond-the-SM extensions predict new particles or interactions directly accessible at the LHC. These signatures range from relatively standard, to those needing special reconstruction algorithms or techniques. This talk will cover several such recent searches at ATLAS.
Speaker: Antti Pirttikoski (Universite de Geneve (CH)) -
70
Autoencoder-based ABCD matrix method for anomaly detection
Anomaly detection has emerged as a promising paradigm for BSM searches at the LHC, leveraging modern machine learning techniques to identify subtle deviations directly in data. In particular, out-of-distribution (OOD) approaches target signals that populate the tails of an anomaly score, offering sensitivity to a wide range of unforeseen signatures. A central challenge in such searches is the estimation of the background in the extreme tail of the distribution: by construction, events in the signal region are poorly modelled by simulation. While extrapolation from control regions is in principle possible, it is typically unreliable, and sideband interpolation is not applicable.
In this talk, I will present a novel strategy that combines OOD anomaly detection with the ABCD method for data-driven background estimation. The approach integrates normalizing flows and variational autoencoders to construct an end-to-end framework that enables robust background predictions directly from data, even in the far tail of the anomaly score distribution. I will present preliminary results demonstrating the performance of the method on a benchmark BSM scenario and discuss its implications for future anomaly-based searches at the LHC.
Speaker: Rafal Maselek -
71
Anomaly detection for multi-jet resonances
The search for physics beyond the standard model is one of the main focuses in high-energy physics. Conventional searches at the LHC, though comprehensive, have not yet shown signs for new physics. Machine learning based anomaly detection has emerged offering a model-agnostic way to enhance the sensitivity of generic searches as compared to those targeting specific signal model. CATHODE (Classifying Anomalies THrough Outer Density Estimation), one of these methods, is a two-step method that combines a data driven background estimation with a classifier flagging potential signal. To date, most studies have mainly focused on dijet resonances.
Extending this approach, we explore signals with multiple decays modes spanning a range of jet multiplicities, leading to a more challenging detection scenario. We demonstrate how a well established idea from jet-substructure physics (recursive soft drop) can be utilized to perform anomaly detection. We present the first application of CATHODE to multi-jet resonances, which enhances the sensitivity beyond the dijet regime and increases the robustness of weakly supervised anomaly detection, thereby broadening its applicability.Speaker: Chitrakshee Yede (Hamburg University (DE)) -
72
Searches for new particles decaying into a quark-antiquark pair at √s=13 TeVSpeaker: Adam Albert Kobert (Rutgers State Univ. of New Jersey (US))
-
68
-
IAC Meeting
-
-
- 73
-
10:00
Coffee Break
-
74
Summary (experiment)Speaker: Hannah Bossi (Massachusetts Inst. of Technology (US))
-
Closing Remarks: Thank You! & BOOST 2027
-
12:00
Lunch Break