Sequential Bayesian inference with correlated heavy-ion datasets
Listen
Radio episode about this paper
Transcript
Introduction to the show: ident: Nuclear Theory. The day in nuclear theory: structure, reactions and the models behind them.
Kai: I'm Kai.
Mira: And I'm Mira, here with Lev.
Kai: We're looking at the paper: "Sequential Bayesian inference with correlated heavy-ion datasets".
Mira: Here's the gist of the paper. Sequential Bayesian inference with correlated heavy-ion datasets investigates how treating datasets as conditionally independent in sequential updating leads to systematic deviations from joint inference when correlations are present The factorized sequential update reproduces the joint posterior only in the limit of conditional independence, and otherwise lead to systematic deviations that grow with the correlation strength The Problem of Factorization The core issue addressed is that when datasets are statistically correlated, treating them as conditionally independent leads to biased posterior estimates because it neglects the joint likelihood structure.
Kai: This approximation becomes inconsistent with the joint likelihood and can lead to biased posterior estimates.
Mira: In realistic applications, different datasets are often correlated due to shared experimental systematics, common model uncertainties, or underlying physical mechanisms that couple multiple observables.
Kai: When such correlations are present, the likelihood no longer factorizes, and sequential updates based on marginal likelihoods do not reproduce the result of a joint analysis.
Title and authors: Kai: So we're looking at this paper now titled "Sequential Bayesian inference with correlated heavy-ion datasets". It tackles how we treat multiple datasets when they aren't independent in a sequence, which is a big issue in experiments where different measurements overlap.
Mira: Exactly. The core idea here is that if you treat those datasets as conditionally independent when you update sequentially, your final result won't match the true joint inference unless there are no correlations at all.
Lev: From my side, this means if we build a real error-correction system based on sequential steps, and the data has correlations—like shared systematic uncertainties—the overhead in our correction is going to be way higher than we assume.
Kai: Right. The paper sets up a controlled environment using pseudo-data with a tunable covariance structure, specifically varying the parameter rho which controls how strong those cross-dataset correlations are between two observable groups.
Mira: It’s interesting because they use this setup to show that factorized sequential updates only reproduce the true joint posterior when the correlation strength, that rho, is zero and you have perfect conditional independence.
Lev: So they're showing us a systematic deviation that grows with the correlation strength if we use those simpler factorized updates instead of following the exact conditional likelihood formulation.
Kai: That’s what they show: these deviations actually grow as you increase the correlation strength between the datasets, which means if you ignore that structure, your posterior estimates get systematically biased.
Mira: And they quantify this impact by using an information decomposition to separate what information is genuinely new from what is just redundant across datasets.
Lev: That idea of separating new versus redundant information helps explain why some parameters are more sensitive to these correlation effects than others in the final results.
Kai: The study points out that parameters like epsilon two and kappa v, which are heavily constrained by flow observables but have dependencies on other sectors through scales like S, show the largest deviations from the true joint result as rho gets stronger.
Mira: That makes sense from a theoretical standpoint because it shows how the structure of parameter sensitivities dictates where the information loss happens most severely in these sequential updates.
Lev: If we were running this on real hardware, knowing that epsilon two is particularly sensitive to these correlations tells us exactly which measurement uncertainties we need to model with higher fidelity during those sequential steps.
Kai: So, what they suggest is that consistent sequential inference demands a proper treatment of correlations right at the likelihood level within a common probabilistic framework.
Mira: It’s about moving away from treating datasets as independent approximations and instead using the exact conditional likelihood when you have these overlapping measurements.
Title and authors: Lev: For us in error correction, it means our priors or intermediate steps need to explicitly account for how data from different sources are coupled before we assume independence to move forward.
Kai: And they show that this approach, the conditional sequential updating, actually stays fully consistent with the joint posterior for Gaussian likelihoods even when correlations are present.
Mira: That contrast is important because it shows that the way you update sequentially matters a lot; not all sequential methods handle correlations equally well.
Lev: So if you're designing an experiment, knowing that conditional updating works consistently gives us a target to aim for in our real-world analysis pipelines.
Kai: To wrap up this part of the paper, they demonstrate how cross-dataset correlations reorganize the information carried by different observables across parameter space.
Mira: It’s anisotropic, meaning it can both suppress and enhance constraints depending on which direction you're looking at in that parameter space.
Lev: That means some parts of the model get tighter constraints while others become fuzzier when those correlations kick in, which is crucial for designing experiments to get the most out of them.
Kai: Overall, this work with "Sequential Bayesian inference with correlated heavy-ion datasets" shows that treating correlated data as independent in a sequence leads to systematic deviations that scale directly with the correlation strength.
Mira: The paper provides a concrete, model-independent assessment of how these correlations affect posterior estimates by comparing joint inference against factorized updates and exact conditional likelihoods.
Lev: If you're trying to make sequential inference robust, this comparison gives you a clear way to test if your approximations are breaking down under realistic data coupling.
Kai: It’s a solid framework for assessing the validity of those prior assumptions we often make in sequential workflows when dealing with multi-observable data.
Mira: So, while the exact formulation is complex, the result is that ignoring correlations at the likelihood level introduces errors that grow with how correlated your datasets actually are.
Lev: We can use this to guide our next steps by knowing exactly where to focus our error modeling efforts when we move from simulation to actual experimental data.
Kai: This paper really highlights how much attention needs to be paid to the structure of the likelihood when you’re dealing with sequential information gathering in complex systems like heavy-ion physics.
Mira: It sets a clear standard for how we should think about combining datasets that are physically coupled, rather than just assuming they are statistically separate pieces of information.
Lev: So if you're building a system where measurements feed into each other sequentially, you need to model the coupling explicitly to avoid these growing systematic errors.
Kai: That’s what we found in "Sequential Bayesian inference with correlated heavy-ion datasets" and it gives us a clear path forward for more rigorous analysis in this area.
The paper's summary: Kai: So, this paper is essentially saying that when you try to figure out parameters sequentially using different datasets, treating those datasets as independent when they are actually correlated leads to systematic errors that get bigger the stronger the correlation is.
Mira: Right. They’re looking at how factorized sequential updates only match the true joint result in one specific case: where there are absolutely no correlations between the data. If there's any coupling, those simpler updates just drift away from reality systematically.
Lev: From an engineering standpoint, that means if we build a real error-correction system based on this sequential method, and our data sources are coupled—like shared systematic uncertainties—the overhead for our correction will scale with how correlated the sources are.
Kai: And they quantify this deviation using something called Kullback–Leibler divergence between the sequential results and the true joint posterior. That number tells us exactly how much information we’re losing when we approximate it this way.
Mira: The decomposition they use is really neat because it splits the data contribution into two parts: what's genuinely new versus what's just redundant across datasets. This redundancy is driven by how much the sensitivities of those different datasets overlap in parameter space.
Lev: That overlap concept is key. It shows that certain parameters, like epsilon two and kappa v, which are primarily constrained by flow measurements but still have links to other sectors via things like the scale S, show the biggest deviations from joint inference when those correlations are present.
Kai: So what this means for us on the ground is that we can’t just assume independence when we're updating our models; we need to explicitly account for that coupling at the likelihood level within our probabilistic framework.
Mira: Exactly. The implication is that consistent sequential inference requires a proper treatment of correlations in the likelihood itself, not just an approximation done after the fact.
Lev: This framework gives us a way to test if our prior assumptions about statistical independence are actually holding up when we look at data that's physically coupled like heavy-ion observables.
Kai: So they’ve shown that cross-dataset correlations reorganize how information is distributed across parameter space, which means some constraints get tighter while others get fuzzier depending on the direction you're looking.
Mira: It’s anisotropic, meaning the effect isn't uniform; it depends entirely on which part of the parameter space you are examining.
Lev: If we were designing an experiment, knowing that parameters like kappa v are especially sensitive to these correlation effects tells us exactly where we need to focus our error modeling efforts during those sequential data-taking steps.
Kai: This paper gives us a clear path forward for how to rigorously assess the assumptions we make when combining multiple datasets in sequential workflows.
Mira: We'll be looking at how this framework can be applied to real heavy-ion data next, seeing if these correlations actually show up in the way they predict.
The paper's improvements: Kai: So, we're looking at how the authors suggest improving their original approach to sequential inference in this paper. Basically, they show that factorized updates are bad when correlations are strong, so the fix is to use a method that respects those dependencies at every step.
Mira: Right. They propose moving towards a conditional sequential update because that actually stays consistent with the true joint posterior even when you have those correlated datasets in your hand. That’s what they show works mathematically for Gaussian likelihoods.
Lev: From an engineering view, that means if we're building a real system, we need to model those correlations explicitly in our sequential steps so the resulting parameter estimates don't systematically drift away from the true answer.
Kai: The paper also points out how you can use that information decomposition they mentioned earlier—separating new and redundant data contributions—to understand *why* things are shifting. It shows how data overlap dictates where the constraints get enhanced or suppressed in parameter space.
Mira: They suggest that this decomposition helps us map out which parts of the model are driven by genuinely new information versus just repeating what we already know from previous measurements. This is crucial for figuring out if a new measurement is actually adding something to our understanding or just confirming old stuff.
Lev: For running this on actual hardware, knowing this overlap structure helps us decide which sensors we need to focus on when the data comes in—we don't want to waste resources modeling redundant information that doesn't add new constraint.
Kai: The implication is that instead of just throwing more data at a problem and hoping it gets better, you need a way to understand the structure of the correlations themselves to know if it’s actually helping or just adding noise in a systematic way.
Mira: So, they're pushing for this kind of structural understanding: realizing that sequential inference isn't just about adding data points; it’s about managing how those data points are coupled together physically.
Lev: That leads us to thinking about future work—if we can nail the methodology here, the next step is testing these principles on more complex models with non-Gaussian likelihoods, which is where the real difficulty lies.
Kai: Exactly. Right now, it’s a nice demonstration with pseudo-data and Gaussian noise, but extending this to real heavy-ion data with more complicated physics would really test how robust this framework is.
Mira: It's a strong foundation because it proves that simply ignoring correlations at the likelihood level is a significant source of error, which means any future analysis needs to be built on top of this structural awareness.
Lev: The next thing we need to see is how much overhead those explicit correlation models introduce when you try to run this on massive datasets that you'd actually collect in an experiment.
Conclusion: Tom: So to wrap up, this paper on "Sequential Bayesian inference with correlated heavy-ion datasets" shows that treating correlated data as independent in sequence introduces systematic errors that grow with how strong those correlations are.
Kai: Basically, it means if you ignore the coupling between your data sources when updating sequentially, your final parameter estimates will just drift away from the true result unless you use a proper conditional likelihood approach.
Mira: The core message is that consistent sequential inference requires accounting for correlations right at the likelihood level within a common probabilistic framework instead of just making an approximation later.
Lev: For us, this tells us that when we're building real hardware and need to run these complex analyses, we have to model those correlations explicitly because ignoring them creates overhead in our error correction steps that scales with the correlation strength.
Kai: It’s a lot of work conceptually, but it gives us a clear way to test if the prior assumptions we make about statistical independence are actually holding up when we look at data that's physically coupled like heavy-ion observables.
Mira: And they show that this structural understanding—separating new versus redundant information—is key to seeing *why* constraints get enhanced or suppressed depending on the parameter direction you're looking at.
Lev: If we can nail this methodology, the next step is testing these principles on more complex models with non-Gaussian likelihoods, which is where the real difficulty lies for our error correction methods.
Kai: So while this work uses pseudo-data to show the concept clearly, it really sets a standard for how we should think about combining multiple datasets in sequential workflows when they're physically coupled.
Mira: It’s a reminder that sequence matters, and you can't just add data points and expect the result to be as good as if they were all independent.
Lev: We need to see how this translates into a practical overhead calculation for real experimental setups, which is where we'll be looking next.
Department of Physics, University of California, Berkeley · Nuclear Science Division, Lawrence Berkeley National Laboratory
nucl-th, hep-ph, nucl-ex
Submitted: 2026-05-18
Updated: 2026-11-12
Comments: v1: 11 pages, 7 figures; v2: Discussion expanded and presentation improved. Main results and conclusions unchanged. Published in Physics Letters B
DOI: 10.1016/j.physletb.2026.141159
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Importance score: 86/100
The gist: Sequential Bayesian inference with correlated heavy-ion datasets investigates how treating datasets as conditionally independent in sequential updating leads to systematic deviations from joint…
Key concepts
- Factorization Problem
- When datasets are correlated, assuming they are independent during sequential updating introduces bias because it ignores the shared physical structure between measurements. This approximation fails to reproduce the true joint posterior unless the data are statistically independent.
- Information Decomposition
- This method separates information from datasets into parts that are genuinely new and parts that overlap or are redundant across different datasets. Correlations cause this information to be redistributed in a parameter-dependent way based on how much each dataset overlaps with others.
- Kullback–Leibler Divergence
- This metric is used to measure the systematic deviation between the posterior distribution derived from a sequential (factorized) update and the true joint posterior. A larger divergence indicates greater information loss or mismatch caused by treating correlated data as independent.
Terminology
Summary
Sequential Bayesian inference with correlated heavy-ion datasets investigates how treating datasets as conditionally independent in sequential updating leads to systematic deviations from joint inference when correlations are present The factorized sequential update reproduces the joint posterior only in the limit of conditional independence, and otherwise lead to systematic deviations that grow with the correlation strength
The Problem of Factorization
The core issue addressed is that when datasets are statistically correlated, treating them as conditionally independent leads to biased posterior estimates because it neglects the joint likelihood structure. This approximation becomes inconsistent with the joint likelihood and can lead to biased posterior estimates. p. 0 In realistic applications, different datasets are often correlated due to shared experimental systematics, common model uncertainties, or underlying physical mechanisms that couple multiple observables. p. 0 When such correlations are present, the likelihood no longer factorizes, and sequential updates based on marginal likelihoods do not reproduce the result of a joint analysis. p. 0
Model Construction and Data Setup
The study employs a reduced model designed to isolate how different classes of observables constrain overlapping parameter directions. The model depends on four parameters, theta = (S, ε2, κN, κv), where S denotes an entropy-like scale and ε2 an eccentricity-like parameter. p. 1 Pseudo-data are generated from a fixed reference parameter point θ⋆ by adding Gaussian fluctuations d = m(θ⋆) + δ, where Σ is the covariance matrix. p. 7 Cross-dataset correlations are introduced through the offdiagonal block Σ12 = ρ Σ˜12 [35], where ρ controls the overall strength of correlations between the two observable groups. p. 2
Information Decomposition
The paper introduces an information decomposition to separate contributions into components that are new and components that are redundant across datasets. Correlations induce a structured, parameter-dependent redistribution of information, governed by the overlap of dataset sensitivities. p. 0 This is quantified by the information flow from D1 to D2, I2←1 = 1/2 log det Σ2 / det Σ21, which quantifies the reduction in uncertainty volume of D2 due to knowledge of D1. p. 4
Impact on Posterior Inference
The comparison between joint inference, factorized sequential updating, and conditional sequential updating shows that factorized updates reproduce the joint posterior only in the limit of conditional independence. Systematic deviations develop between the factorized sequential result and the joint posterior as cross-dataset correlations increase. p. 6 The magnitude of these deviations is quantitatively controlled by the mismatch between marginal and conditional information, which can be measured using Kullback–Leibler divergence. p. 1
Conclusion and Implications
The results demonstrate that cross-dataset correlations reorganize the information carried by different observables, modifying how constraints are distributed across parameter space. The effect is anisotropic, with both suppression and enhancement of constraints along different parameter directions. p. 5 Consistent sequential inference requires a proper treatment of correlations at the likelihood level within a common probabilistic framework. p. 7
Summary and Conclusions
In this work, we have investigated sequential Bayesian inference in the presence of correlated datasets within a controlled framework. We have shown that factorized sequential updates depart increasingly from the joint result as the strength of cross-dataset correlations grows, while conditional updating remains fully consistent with the joint posterior. p. 6 These deviations can be understood in terms of an information decomposition, in which correlations partition the contribution of a dataset into components that are genuinely new and components that are not independent of previously incorporated data. p. 6 Consistent sequential inference therefore requires a proper treatment of correlations at the likelihood level within a common probabilistic framework. p. 7 The framework developed here provides a way to assess the validity of this assumption and to quantify its impact. p. 7
Acknowledgements
The author acknowledges valuable discussions with S. Jaiswal, P. p. 7 M. Jacobs, and other members of the JETSCAPE Collaboration. p. 7 This work was supported in part by the U. p. 7S. Department of Energy, Office of Science, Office of Nuclear Physics under Grant No. p. 7 DE-AC02-05CH11231. p. 7 The author acknowledges the use of ChatGPT for assistance with grammar refinement and clarity improvement during manuscript preparation. p. 7
Appendix A
In this appendix, we provide the explicit construction of the reduced model used in the main text. The purpose of this model is not to reproduce realistic heavy-ion data, but to provide a controlled and transparent setting in which different classes of observables probe partially overlapping aspects of a small set of parameters. p. 7 The guiding principle is to retain the minimal structure necessary to capture multi-observable inference, while keeping the model simple enough to allow direct evaluation of the likelihood without introducing additional complications such as emulators or surrogate models. p. 7
Appendix B
To clarify how correlations propagate from the data level to parameter inference, we examine the Gaussian likelihood in block form. The goal is to make explicit how the covariance structure modifies the information carried by different datasets, and how this modification is encoded in both the likelihood and its parameter-space representation. p. 8 The joint Fisher matrix can therefore be written as Fjoint = F1 + F21. p. 9 This decomposition provides the parameter-space counterpart of the likelihood factorization above: the total information is the sum of the contribution from D1 and the independent component of D2. p. 9 The off-diagonal structure of Σ−1 determines how correlations are projected into parameter space locally, while the induced reshaping of the likelihood governs how these effects accumulate to produce the global posterior distortions discussed in the main text. p. 9
References
[1] R. Trotta, Bayes in the sky: Bayesian inference and model selection in cosmology, Contemp. p. 9 Phys. 49 (2008) 71–104
[2] D. MacKay, Information Theory, Inference and Learning Algorithms, Cambridge University Press, 2003
[3] A. R. Liddle, Statistical methods for cosmological parameter selection and estimation, Ann. p. 9 Rev. Nucl. Part. Sci. 59 (2009) 95–114
[4] S. Pratt, E. Sangaline, P. p. 9 Sorensen, H. p. 9 Wang, Constraining the Eq of State of Super-Hadronic Matter from Heavy-Ion Collisions, Phys. p. 9 Rev. Lett. 114 (2015) 202301
[5] J. E. Bernhard, J. p. 9 S. Moreland, S. p. 9 A Bass, J Liu, U Heinz, Applying Bayesian parameter estimation to relativistic heavy-ion collisions: simultaneous characterization of the initial state and quark-gluon plasma medium, Phys. p. 9 Rev. C 94 (2) (2016) 024907
[6] J. E. Bernhard, J S Moreland, S A Bass, Bayesian estimation of the specific shear and bulk viscosity of quark–gluon plasma, Nature Phys. p. 9 15 (11) (2019) 1113–1117
[7] G. Nijs, W van der Schee, U Gursoy, R Snellings, Transverse Momentum Differential Global Analysis of Heavy-Ion Collisions, Phys. p. 9 Rev. Lett. 126 (20) (2021) 202301
[8] G. Nijs, W van der Schee, U Gursoy, R Snellings, Bayesian analysis of heavy ion collisions with the heavy ion computational framework Trajectum, Phys. p. 9 Rev. C 103 (5) (2021) 054909
[9] D. Everett, et al. p. 9 p. 9, Phenomenological constraints on the transport properties of QCD matter with data-driven model averaging, Phys. p. 9 Rev. Lett. 126 (24) (2021) 242301
[10] D. Everett, et al., Multisystem Bayesian constraints on the transport coefficients of QCD matter, Phys. p. 9 Rev. C 103 (5) (2021) 054904
[11] W Fan, et al., New metric improving Bayesian calibration of a multistage approach studying hadron and inclusive jet suppression, Phys. p. 9 Rev. C 109 (6) (2024) 064903
[12] S Cao, et al., Determining the jet transport coefficient q from inclusive hadron suppression measurements using Bayesian parameter estimation, Phys. p. 9 Rev. C 104 (2) (2021) 024905
[13] R Ehlers, et al., Bayesian inference analysis of jet quenching using inclusive jet and hadron suppression measurements, Phys. p. 9 Rev. C 111 (5) (2025) 054913
[14] A Mankolli, et al., Longitudinal Dynamics of Large and Small Systems from a 3D Bayesian Calibration of RHIC Top-energy Collision Data arXiv:2601. p. 917234
[15] L Du, Multi-observable analysis of jet quenching using Bayesian inference with JETSCAPE, Nucl. Phys. A 1060 (2025) 123100
[16] D S Sivia, Data analysis: a Bayesian tutorial, second edition. Edition, Oxford science publications, Clarendon, Oxford, 2006
[17] U von Toussaint, Bayesian inference in physics, Rev. Mod. Phys. 83 (2011) 943–999
[18] J.-F. Paquet, Applications of emulation and Bayesian methods in heavyion physics, J. p. 0 Phys. G 51 (10) (2024) 103001
[19] C Bishop, Pattern Recognition and Machine Learning, Information Science and Statistics, Springer, 2006
[20] R Van de Schoot, S Depaoli, R King, B Kramer, K Martens, M G Tadesse M Vannucci A Gelman D Veen J Willemsen et al., Bayesian statistics and modelling, Nature Reviews Methods Primers 1 (1) (2021) 1
[21] H Roch, C Shen, Learning informed prior distributions with normalizing flows for Bayesian analysis, Phys. Rev. C 113 (3) (2026) 034903
[22] P Bozek, Transverse-momentum–flow correlations in relativistic heavyion collisions, Phys. Rev. C 93 (4) (2016) 044908
[33] L Du, A Sorensen, M Stephanov, The QCD phase diagram and Beam Energy Scan physics: A theory overview, Int J Mod Phys E 33 (07) (2024) 2430008
[35] R A Soltz, D A Hangal, A Angerami, Simple model to investigate jet quenching and correlated errors for centrality-dependent nuclear modification factors in relativistic heavy-ion collisions, Phys. Rev. C 111 (3) (2025) 034911
[36] U Heinz, R Snellings, Collective flow and viscosity in relativistic heavyion collisions, Ann. Rev. Nucl. Part. Sci. 63 (2013) 123–151
[40] B Alver, G Roland, Collision geometry fluctuations and triangular flow in heavy-ion collisions, Phys. Rev. C 81 (2010) 054905
[41] F G Gardim, F Grassi, M Luzum, J Y Ollitrault Mapping the hydrodynamic response to the initial geometry in heavy-ion collisions, Phys. Rev. C 85 (2012) 024908
[42] K Aamodt, et al., Centrality dependence of the charged-particle multiplicity density at mid-rapidity in Pb-Pb collisions at √sNN = 2. p. 776 TeV, Phys. p. 7 p. 7 Rev. Lett. 106 (2011) 032301
[43] K Aamodt, et al., Higher harmonic anisotropic flow measurements of charged particles in Pb-Pb collisions at √sNN=2. p. 076 TeV, Phys. Rev. Lett. 107 (2011) 032301
[44] C E Rasmussen, C K I Williams, Gaussian Processes for Machine Learning, The MIT Press, 2005
[45] S Kullback, R A Leibler, On information and sufficiency, The annals of mathematical statistics 22 (1) (1951) 79–86
[46] J H Putschke, et al., The JETSCAPE framework arXiv:1903. p. 6
Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.
Improvements for AI systems
-
Improved AI systems can diagnose inconsistencies in sequential Bayesian inference by comparing factorized updates against joint inference when datasets are correlated, as shown by
factorized updates reproduce the joint posterior only in the limit of conditional independence, and otherwise lead to systematic deviations that grow with the correlation strength.
-
The improved system can quantify the impact of dataset correlations by calculating Kullback–Leibler divergence between sequential posteriors and joint posteriors, which measures
the information loss incurred when the joint posterior is approximated by a sequential construction.
-
The system can identify parameter-dependent information redistribution by introducing an information decomposition that separates contributions into
components that are new and components that are redundant across datasets,
linking these to theoverlap of dataset sensitivities.
-
The AI can predict which parameters are most sensitive to correlation effects, as demonstrated by observing that
parameters such as ε2 and κv, which are primarily constrained by the flow observables (D2) but remain indirectly coupled to D1 through shared dependencies (notably via S), exhibit the largest deviations.
-
The system can assess the validity of prior assumptions in sequential workflows by testing them against data, specifically quantifying
the impact of priors encodes information from previous datasets and implicitly assume statistical independence from the data incorporated at later stages.
Sources
- Bayes in the sky: Bayesian inference and model selection in cosmology
- Statistical methods for cosmological parameter selection and estimation
- Constraining the Eq. of State of Super-Hadronic Matter from Heavy-Ion Collisions
- Applying Bayesian parameter estimation to relativistic heavy-ion collisions: simultaneous characterization of the initial state and quark-gluon plasma medium
- A transverse momentum differential global analysis of Heavy Ion Collisions
- Bayesian analysis of heavy ion collisions with the heavy ion computational framework Trajectum
- Phenomenological constraints on the transport properties of QCD matter with data-driven model averaging
- Multi-system Bayesian constraints on the transport coefficients of QCD matter
- A new metric improving Bayesian calibration of a multistage approach studying hadron and inclusive jet suppression
- Determining the jet transport coefficient $\hat{q}$ from inclusive hadron suppression measurements using Bayesian parameter estimation
- Bayesian Inference analysis of jet quenching using inclusive jet and hadron suppression measurements
- Longitudinal Dynamics of Large and Small Systems from a 3D Bayesian Calibration of RHIC Top-energy Collision Data
- Applications of emulation and Bayesian methods in heavy-ion physics
- Learning Informed Prior Distributions with Normalizing Flows for Bayesian Analysis
- Transverse momentum-flow correlations in relativistic heavy-ion collisions
- Extracting the shear viscosity of the quark-gluon plasma from flow in ultra-central heavy-ion collisions
- Rapidity scan with multistage hydrodynamic and statistical thermal models
- Triangularity and Dipole Asymmetry in Heavy Ion Collisions
- Estimating initial state and quark-gluon plasma medium properties using a hybrid model with nucleon substructure calibrated to $p$-Pb and Pb-Pb collisions at $\sqrt{s_\mathrm{NN}}=5.02$ TeV
- Advancing multimessenger approaches in heavy-ion collisions: Insights from electromagnetic probes
Related papers
- Constrained Hamiltonian Formulation of Kinetic Theory for Relativistic Spin- 1/2 Particles
- Active-subspace learning reveals dominant parameter directions in nuclear observables
- Long- and short-range longitudinal flow decorrelations in the quark-gluon plasma
- Recurrent Fourier Neural Operator for Event-by-Event Relativistic Hydrodynamics
- Bayesian Inference of the Landau Parameter G'0 from Joint Gamow-Teller Measurements
- UrQMD-Agent: An LLM-Assisted Framework for Microscopic Tracing of Proton Azimuthal Correlations