Pith. sign in

REVIEW 4 major objections 6 minor 36 references

Towards Foundation Models for Experimental Readout Systems Combining Discrete and Continuous Data

T0 review · 4 major / 6 minor · reviewed 2026-08-15 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read A single transformer with split pixel/time vocabularies can generate Cherenkov hit patterns for a future EIC detector and transfer to pion/kaon identification and noise filtering.

desk verdict A serious methods paper with a genuinely new split-vocabulary attention scheme for detector readout; the architecture is worth knowing, but the 'high-fidelity' claim is overstated at the kinematic point where it matters most. read the letter →

arxiv 2505.08736 v2 pith:GRTLJF5R submitted 2025-05-13 cs.LG hep-exnucl-exphysics.ins-det

James Giroux , Cristiano Fanelli This is my paper · ORCID
classification cs.LGhep-exnucl-exphysics.ins-det
keywords foundationmodelsCherenkovdetectorsnext-tokenpredictioncausalcross-attentionmixtureofexpertsparticleidentificationfastsimulationnoisefiltering
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper argues that one autoregressive transformer, trained with separate token vocabularies for pixel location and arrival time, can act as a proto-foundation model for pixelated detector readouts at the future Electron-Ion Collider. The core proposal is to avoid lossy learned tokenization: keep native pixel indices as discrete tokens, bin continuous arrival times at a fraction of the sensor resolution, and fuse the two streams with causal multi-head cross-attention where each time token queries its plausible pixel locations. Kinematic conditions, momentum magnitude and polar angle, are prepended as continuous embeddings, and a class-conditioned mixture of experts lets a single model generate both pion and kaon patterns. Closure tests on the High Performance DIRC show near-unity spatial ratios, photon yields that track the reference simulation across polar angle, pion/kaon separation meeting the 3-sigma requirement at 6 GeV/c, and token-level noise filtering with roughly 90 percent average precision. If the approach holds, detector simulation and reconstruction for Cherenkov systems could share one trainable backbone.

What carries the argument

The load-bearing mechanism is the split-vocabulary causal sequence with Causal Multi-Head Cross-Attention (CMHCA): time embeddings are the query stream and pixel embeddings the key/value stream, so each arrival time queries its plausible pixel locations while the mask preserves autoregressive causality. Two linear heads then predict the next token over each vocabulary, avoiding any joint space-time vocabulary of roughly 36 million tokens; the model instead learns 6,144 pixel tokens and 5,920 time tokens and combines them by attention. $\ell^2$-normalized query and key matrices with a learned scale, plus pre-layer normalization, keep attention focused rather than collapsed onto the most populated configurations across phase space. For class-conditional generation, the transformer's feed-forward networks are replaced by a mixture of experts with fixed class routing, so pion and kaon sequences pass through different expert subsets inside one shared backbone.

What would settle it

Regenerate the published closure-test kinematics with temperature T=1.0 and nucleus probability p=1.0, so the sampler draws directly from the trained distribution, and compare the joint pixel-time distribution to the reference simulation; a sharp degradation there would show that the reported fidelity depends on hand-tuned sampling rather than on a learned joint distribution.

Watch

Extended reading notes

Core claim

The central claim is that next-token prediction over a time-sorted sequence of detector hits is a viable generative target for Cherenkov readout at detector-level resolution, provided space and time live in separate vocabularies and are fused by causal cross-attention. The pixel vocabulary is the native detector readout, 6,144 indices; time is linearly binned at one-quarter of the sensor resolution, giving 5,920 tokens. The model predicts the next pixel and the next time jointly, with two cross-entropy losses, conditioned on prepended continuous embeddings of momentum magnitude and polar angle. Replacing the feed-forward layers with class-routed mixture-of-experts layers produces generations that the paper finds in strong agreement with independent per-class models, so pions and kaons can share one model. Closure metrics show spatial distributions in ratio near one and photon yields in good agreement, while separation power reaches the 3-sigma EIC requirement at 6 GeV/c but remains below the reference simulation and the normalizing-flow baseline; the paper attributes the gap to inherent smoothing and to the freedom in choosing generation parameters.

Load-bearing premise

That ordering a track's Cherenkov photons by time and predicting them one token at a time is a well-posed way to learn the joint hit distribution, even though the order of individual photons is not physically meaningful; the paper itself calls next-token prediction of Cherenkov photons 'highly ambiguous' and relies on hand-tuned sampling temperature and nucleus probability (T=1.05, p=0.995) to keep generations coherent.

Editorial extensions

If this is right

  • A single class-conditioned model can generate multiple particle species in one backbone, since closure tests show MoE generations agree with independent per-class models.
  • Because the spatial vocabulary is the native pixel readout, generated samples preserve per-pixel geometric structure, including kaleidoscopic patterns near the expansion volume that a continuous normalizing-flow generator smooths away.
  • The same pretrained backbone fine-tunes into a pion/kaon classifier with faster convergence and slightly higher accuracy than training from scratch.
  • Token-level noise filtering reaches about 0.90 average precision and about 0.99 rejection AUC, so the readout model can also tag dark-rate noise hits in high-rate EIC environments.
  • Directly learning sequence length removes the need for post-hoc photon-yield modeling as a function of phase space, a step previous fast Cherenkov simulation required.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A stronger test than the published marginals would compare full joint pixel-time correlations between generated and reference tracks; divergence there would indicate the success rests on tuned sampling, not on a faithful learned joint distribution.
  • An ablation that randomizes hit order within tracks during training would directly probe whether the autoregressive factorization is learning the hit set or the dataset's specific ordering; if it collapses, the ordering assumption, not the architecture, is carrying the result.
  • If the split-vocabulary pattern transfers, a detector stream could chain time, energy, and space through sequential cross-attention blocks, which the paper sketches as future work and which would extend the method to calorimeters without VQ-VAE resolution loss.
  • The low-momentum PID advantage suggests a momentum-weighted loss could reallocate capacity toward the high-momentum region where pion/kaon separation is hardest; the paper lists this as future work without testing it.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 6 minor

Summary. The manuscript proposes an autoregressive transformer for generating low-level Cherenkov detector readouts—pixel indices and arrival times—for the High Performance DIRC at the future EIC. The architecture introduces four claimed innovations: separate discrete vocabularies for space and time fused through Causal Multi-Head Cross-Attention (CMHCA), continuous kinematic conditioning via prepended context embeddings, fine-grained continuous-variable tokenization without joint vocabulary inflation, and class-conditional generation via a Mixture of Experts (MoE). The model is trained on Geant4-generated track-level hit sequences and evaluated through cumulative distribution ratio plots, photon-yield comparisons, and FastDIRC KDE separation power at fixed momenta, with additional downstream tasks of pion/kaon identification and noise filtering. The central claim is that a single (proto) foundation model can generate kinematically conditioned Cherenkov hit patterns with high fidelity and transfer to reconstruction tasks.

Significance. If the approach holds, it offers a useful step toward fast, conditional simulation of pixelated Cherenkov detectors without the resolution loss of VQ-VAE tokenization, while sharing a single backbone across particle classes and downstream tasks. The split-vocabulary CMHCA idea is interesting and plausibly transferable to other detector systems. The paper is honest in places: it publicly releases code, compares against Geant4 and a normalizing-flow baseline, and explicitly acknowledges some performance degradation. However, the main quantitative evidence for generation fidelity is mixed: at 6 GeV/c the transformer variants achieve KDE separation of 2.55–2.59 s.d., below both the NF baseline (2.91 s.d.) and Geant4 (3.72 s.d.), and the sampling hyperparameters appear to be selected on the target distributions without a sensitivity analysis. The significance is therefore moderate and contingent on addressing the open generation-fidelity and robustness questions.

major comments (4)
  1. [Section 4, KDE based Evaluation (Fig. 6, bottom)] At |p|=6 GeV/c the transformer variants reach 2.55–2.59 s.d. separation, which is below both the Normalizing Flow baseline (2.91 s.d.) and Geant4 (3.72 s.d.). This contradicts the abstract's 'high-fidelity generation' characterization. Please either soften the central claim, provide a quantitative argument for why this gap is acceptable (e.g., systematic uncertainty or downstream physics impact), or add additional metrics that support the high-fidelity claim. The text acknowledges degradation, but the abstract and conclusions do not reflect it, and this discrepancy is central to the paper's main claim.
  2. [Section 4, Generative Model Evaluation] The sampling parameters T=1.05 and p=0.995 are explicitly selected because they 'have shown to provide the most consistent cumulative distributions across the phase-space', and the paper states that small changes can lead to significantly different outputs. Because the cumulative distribution ratios are the primary quantitative closure evidence, the reported agreement may be a result of tuning rather than a learned joint distribution. Please provide a sensitivity scan over T and p, report how the cumulative ratios and the KDE separation power vary, and specify a validation protocol that avoids selecting sampling parameters on the evaluation distributions.
  3. [Section 4, Generative Model Evaluation (next-token ambiguity)] The paper states that 'the idea of next token prediction of Cherenkov photons is highly ambiguous', yet the model is trained as an autoregressive next-token predictor over time-sorted hit sequences. Since the physical hit set is permutation-invariant, the factorization imposes a conditional structure that may not be unique or well-posed. Please justify why the learned conditional distribution over the sorted sequence correctly represents the joint hit distribution, and provide evidence (e.g., permutation-invariance or set-level generation metrics) that the generation quality is robust to the ordering convention and to time-bin tie-breaking.
  4. [Section 3, Methods (split vocabularies and CMHCA)] The four claimed innovations are not isolated by ablation. For example, no comparison is shown against a single-joint-vocabulary transformer, a VQ-VAE tokenization baseline, or a model without CMHCA under otherwise identical training conditions. Without these ablations, the paper does not establish which components are responsible for the reported behaviour. Please add ablations or explicitly temper the innovation claims to what the current comparisons support.
minor comments (6)
  1. [Section 4, Histogram Level Evaluations] The cumulative distribution ratios in Fig. 4 (and Appendices A/B) are shown without statistical uncertainties; please add uncertainty bands or state that the deviations are within statistical noise.
  2. [Section 4, Particle Identification / Fig. 8] The momentum labels '3 Gev/c' and '6 Gev/c' contain a typo; they should read '3 GeV/c' and '6 GeV/c'.
  3. [Section 5, Summary and Conclusions] The statement that the model 'achieves the desired separation power (3σ at 6 GeV/c)' refers to the PID task (Fig. 8), but at generation time the KDE separation in Fig. 6 is below 3σ at 6 GeV/c; please clarify which result is being cited.
  4. [Section 4, Noise Filtering] The sentence 'we see our model is able to efficiently able to characterize detector noise' contains a duplicated 'able to'; please correct.
  5. [Section 4, Geometric Effects] The claim that the autoregressive model 'more effectively captures the kaleidoscopic effect' is based on a single visual comparison in Fig. 7; please either quantify this effect or present it as qualitative evidence.
  6. [Abstract and Introduction] The abstract and introduction emphasize 'fast' generation, but no runtime or computational-cost comparison with Geant4 or NF is reported; please add a timing measurement or remove the speed claim.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the model's generation targets are evaluated against Geant4 ground truth and the FastDIRC KDE metric, which are independent of the model's construction.

full rationale

The central derivation is a conditional autoregressive model for Cherenkov photon pixel and time sequences, trained on Geant4-simulated hpDIRC data from the authors' prior dataset paper [18] and evaluated against Geant4 reference distributions and the FastDIRC KDE metric [28]. The generated quantity is the empirical hit distribution, which is not used to define the model's architecture, vocabularies, CMHCA mechanism, kinematic context embeddings, or MoE routing. The quantitative closure tests use cumulative distribution ratios, photon-yield comparisons, and pion/kaon separation power, all compared with Geant4 and with Normalizing Flow/DLL baselines; these are external references rather than outputs of the proposed model. Reusing [18] for the dataset and the KDE evaluation procedure is a self-citation, but it is not load-bearing in a circular way: the Geant4 ground truth is produced by an independent physics simulation, the FastDIRC KDE comes from [28], and no claimed result is defined in terms of a parameter fitted to the evaluation targets. The paper's own admission that the transformer variants degrade at 6 GeV/c relative to Geant4 and NF (Fig. 6) weakens the high-fidelity claim but is a correctness/evidence-strength issue, not a circularity. The selection of sampling parameters (T=1.05, p=0.995) on the phase-space used for evaluation is a methodological limitation that may inflate apparent agreement and leave the result sensitive to hyperparameters, but it does not make the generated distribution equivalent to the input by construction. The acknowledged ambiguity of next-token ordering for Cherenkov photons is likewise a modeling concern rather than a circular step. No equation in the paper reduces to its own input, no fitted parameter is renamed as a prediction, and no uniqueness claim is imported from the authors' prior work. Therefore no circular step is identified.

Assumptions & free parameters 4 free parameters · 4 assumptions · 0 invented entities

No new physical entities are introduced. The central claim rests on the validity of the simulated dataset, the autoregressive ordering of hits, the choice of continuous-time quantization, and the hand-tuned sampling parameters that make generations match the reference distributions.

free parameters (4)
  • Sampling temperature T = 1.05
    Chosen in Section 4 to make generated cumulative distributions best match Geant4 ground truth; not learned and not reported with sensitivity analysis.
  • Nucleus sampling p = 0.995
    Chosen alongside temperature to improve distribution agreement; generation quality depends on this hand-tuned value.
  • Time binning resolution = 1/4 of intrinsic timing resolution
    Discretization granularity for continuous time; authors note 1/10 had negligible impact, so not strongly load-bearing, but still a modeling choice.
  • Max sequence length = 250 hits
    Truncates tracks; affects observed photon yield for high-multiplicity tracks and is a free choice.
assumptions (4)
  • domain assumption Autoregressive factorization over time-sorted hit sequences is a valid target for learning the joint hit distribution.
    The paper itself calls next-token prediction highly ambiguous for Cherenkov photons (Section 4).
  • domain assumption The Geant4-based dataset from Giroux et al. [18] faithfully represents hpDIRC detector response.
    All training and evaluation rests on this simulated dataset; no real EIC detector data exists yet.
  • domain assumption FastDIRC KDE separation power is a valid physics metric for fast-simulation fidelity.
    Used as the main quantitative closure test; metric is inherited from [18].
  • ad hoc to paper Causal cross-attention with time as query and space as key/value preserves the joint spatio-temporal distribution.
    The architecture choice is motivated by one-to-many mappings but not proven to be lossless; it is a design assumption.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Towards Foundation Models for Experimental Readout Systems Combining Discrete and Continuous Data." pith.science (2026). https://pith.science/paper/GRTLJF5R

@misc{pith2026250508736,
  author       = {Pith},
  title        = {Pith review of: Towards Foundation Models for Experimental Readout Systems Combining Discrete and Continuous Data},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/GRTLJF5R}},
  note         = {Machine review of arXiv:2505.08736}
}
read the original abstract

We present a (proto) Foundation Model for Nuclear Physics, capable of operating on low-level detector inputs from Imaging Cherenkov Detectors at the future Electron Ion Collider. Building upon established next-token prediction approaches, we aim to address potential challenges such as resolution loss from existing tokenization schemes and limited support for conditional generation. We propose four key innovations: (i) separate vocabularies for discrete and continuous variates, combined via Causal Multi-Head Cross-Attention (CMHCA), (ii) continuous kinematic conditioning through prepended context embeddings, (iii) scalable and simple, high-resolution continuous variate tokenization without joint vocabulary inflation, and (iv) class conditional generation through a Mixture of Experts. Our model enables fast, high-fidelity generation of pixel and time sequences for Cherenkov photons, validated through closure tests in the High Performance DIRC. We also show our model generalizes to reconstruction tasks such as pion/kaon identification, and noise filtering, in which we show its ability to leverage fine-tuning under specific objectives.

Figures

Figures reproduced from arXiv: 2505.08736 by the authors.

Figure 1
Figure 1. Optical box output: Individual tracks leave sparse hit patterns (red points) integrated over time on the hpDIRC readout plane. The denser hit pattern is obtained by accumulating multiple tracks with the same kinematics. Figure and caption taken from [18]. 3. Methods Tokenization of Space and Time Given the pixelated nature of the readout system, no additional tokenization is required along the spatial dimension; eac… view at source ↗
Figure 2
Figure 2. Architecture: Spatial and temporal sequences are first independently embedded via learnable projections from their respective vocabularies, with positional encodings added along each dimension. Kinematic information (momentum magnitude and polar angle) is projected from continuous space and prepended as contextual tokens to both sequences. The time embeddings serve as queries (Q), while the spatial embeddings provid… view at source ↗
Figure 3
Figure 3. Fast Simulation at 6 GeV/c: Fast Simulation of kaons (left column of plots), and pions (right column of plots) at 6 GeV/c at various polar angles, using Nucleus sampling, fixed temperature and a mixture of four experts (class conditional routing). generations, we provide cumulative distributions in [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: Ratio of synthetic to original data distributions at 6 GeV/c: Ratio plots for kaons (left column) and pions (right column), comparing generations to original data. The top row shows results from independent models using nucleus sampling with fixed temperature; the midd…
Figure 5
Figure 5. Figure 5: Photon Yield Comparison at 6 GeV/c: Comparison of generated photon yield for kaons (left column) and pions (right column), using Nucleus sampling and fixed temperature for independent models (top row), a combined model with two experts (middle row) and a combined model…
Figure 6
Figure 6. Figure 6: Fidelity Studies: The fidelity of the generated samples is assessed via pion–kaon separation power, using the FastDIRC KDE-based metric. Comparisons are shown between Geant4 reference data and synthetic populations generated via fast simulation methods, including Norma…
Figure 7
Figure 7. Figure 7: Geometrical Effect Comparison at 6 GeV/c: Comparison of (a) Normalizing Flows and (b) our autoregressive model, with (b) more effectively capturing the kaleidoscopic effect. 9We have implemented and tested a wide variety of sampling methods, see https://github. com/wmd…
Figure 8
Figure 8. Figure 8: Particle Identification Performance Comparison: PID performance of pions and kaons using (a) Normalizing Flow Delta-Loglikelihood (NF-DLL) method, and our method at 3 GeV/c (top) and 6 GeV/c (bottom) for various values of the polar angle, and (b) comparison of accuracy…
Figure 9
Figure 9. Figure 9: Noise Filtering at 6 GeV/c: Noise filtering performance of our method at 6 GeV for (a) kaons and (b) pions, shown as precision-recall curves and noise rejection versus signal efficiency. The precision-recall curves are insensitive to class imbalance. The dark rate cont…
Figure 10
Figure 10. Figure 10: (right) shows the validation metrics, namely the precision, recall and F1 score for both fine-tuned and models trained from scratch as a function of training iteration. 0 10 20 30 40 50 60 70 Training Iterations 0.0085 0.0090 0.0095 0.0100 0.0105 0.0110 ×10 4 Fine-tun…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

36 extracted references · 18 canonical work pages

  1. [1]

    Finke T, Kr¨ amer M, M¨ uck A and T¨ onshoff J Learning the language of QCD jets with transformers 2023 Journal of High Energy Physics 2023 1–18

    work page 2023

  2. [2]

    Birk J, Hallin A and Kasieczka G OmniJet-α: the first cross-task foundation model for particle physics 2024 Machine Learning: Science and Technology 5 035031

    work page 2024

  3. [3]

    Mikuni V and Nachman B OmniLearn: A method to simultaneously facilitate all jet physics tasks 2024 arXiv preprint arXiv:2404.16091

    arXiv 2024

  4. [4]

    Bardhan J, Agrawal R, Tilak A, Neeraj C and Mitra S Hep-jepa: A foundation model for collider physics ICLR 2025 Workshop on World Models: Understanding, Modelling and Scaling

    2025

  5. [5]

    Leigh M, Klein S, Charton F, Golling T, Heinrich L, Kagan M, Ochoa I and Osadchy M Is tokenization needed for masked particle modelling? 2025 Machine Learning: Science and Technology

    2025

  6. [6]

    Vigl M, Hartman N and Heinrich L Finetuning foundation models for joint analysis optimization in High Energy Physics 2024 Machine Learning: Science and Technology 5 025075

    work page 2024

  7. [7]

    Golling T, Heinrich L, Kagan M, Klein S, Leigh M, Osadchy M and Andrew Raine J Masked particle modeling on sets: towards self-supervised high energy physics foundation models 2024 Machine Learning: Science and Technology 5 035074

    work page 2024

  8. [8]

    Bumblebee: Foundation Model for Particle Physics Discovery 2024 arXiv preprint arXiv:2412.07867

    Wildridge A J, Rodgers J P, Colbert E M, Jung A W, Liu M et al. Bumblebee: Foundation Model for Particle Physics Discovery 2024 arXiv preprint arXiv:2412.07867

    arXiv 2024

Show all 36 references
  1. [9]

    Harris P, Kagan M, Krupa J, Maier B and Woodward N Re-simulation-based self-supervised learning for pre-training foundation models 2024 arXiv preprint arXiv:2403.07066

    2024 arXiv

  2. [10]

    Huang A, Melkani Y, Calafiura P, Lazar A, Murnane D T, Pham M T and Ju X A language model for particle tracking 2024 arXiv preprint arXiv:2402.10239

    2024 arXiv

  3. [11]

    Core 8 026

    Butter A, Huetsch N, Schweitzer S P, Plehn T, Sorrenson P and Spinner J Jet diffusion versus JetGPT – Modern networks for the LHC 2025 SciPost Phys. Core 8 026

    2025

  4. [12]

    Peebles W and Xie S 2023 Scalable diffusion models with transformers Proceedings of the IEEE/CVF international conference on computer vision pp 4195–4205

    2023

  5. [13]

    Birk J, Gaede F, Hallin A, Kasieczka G, Mozzanica M and Rose H OmniJet- {αC}: Learning point cloud calorimeter simulations using generative transformers 2025 arXiv preprint arXiv:2501.05534

    2025 arXiv

  6. [14]

    Neural discrete representation learning 2017 Advances in neural information processing systems 30

    Van Den Oord A, Vinyals O et al. Neural discrete representation learning 2017 Advances in neural information processing systems 30

    2017

  7. [15]

    Lin H, Cheng X, Wu X and Shen D 2022 CAT: Cross Attention in Vision Transformer 2022 IEEE international conference on multimedia and expo (ICME) (IEEE) pp 1–6

    2022

  8. [16]

    Fei J, Li D, Deng Z, Wang Z, Liu G and Wang H Video-ccam: Enhancing video-language understanding with causal cross-attention masks for short and long videos 2024 arXiv preprint arXiv:2408.14023

    2024 arXiv

  9. [17]

    Fanelli C, Giroux J and Stevens J Deep (er) reconstruction of imaging Cherenkov detectors with swin transformers and normalizing flow models 2025 Machine Learning: Science and Technology 6 015028

    2025

  10. [18]

    Giroux J, Martinez M and Fanelli C 2025 Generative Models for Fast Simulation of Cherenkov Detectors at the Electron-Ion Collider (arXiv: 2504.19042)

    2025 arXiv

  11. [19]

    Ameli F, Battaglieri M, Berdnikov V V, Bond´ ı M, Boyarinov S, Brei N, Celentano A, Cappelli L, Chiarusi T, De Vita R et al. Streaming readout for next generation electron scattering experiments 2022 The European Physical Journal Plus 137 958 Towards Foundation Models for Expe...

    2022

  12. [20]

    Vaswani A, Shazeer N, Parmar N, Uszkoreit J, Jones L, Gomez A N, Kaiser L and Polosukhin I Attention is all you need 2017 Advances in neural information processing systems 30

    2017

  13. [21]

    Hendrycks D and Gimpel K Gaussian error linear units (GeLUs) 2016 (arXiv: 1606.08415)

    2016 arXiv

  14. [22]

    Ba J L, Kiros J R and Hinton G E Layer normalization 2016 arXiv preprint arXiv:1607.06450

    2016 arXiv

  15. [23]

    Xiong R, Yang Y, He D, Zheng K, Zheng S, Xing C, Zhang H, Lan Y, Wang L and Liu T 2020 On layer normalization in the transformer architecture International conference on machine learning (PMLR) pp 10524–10533

    2020

  16. [24]

    Henry A, Dachapally P R, Pawar S and Chen Y Query-key normalization for transformers 2020 (arXiv:2010.04245)

    2020 arXiv

  17. [25]

    Lin T Y, Goyal P, Girshick R, He K and Doll´ ar P 2017 Focal loss for dense object detection Proceedings of the IEEE international conference on computer vision pp 2980–2988

    2017

  18. [26]

    Shazeer N, Mirhoseini A, Maziarz K, Davis A, Le Q, Hinton G and Dean J Outrageously large neural networks: The sparsely-gated mixture-of-experts layer 2017 arXiv preprint arXiv:1701.06538

    2017 arXiv

  19. [27]

    Genovese D, Sgroi A, Devoto A, Valentine S, Wood L, Sebastiani C, Giagu S, D’Onofrio M and Scardapane S Mixture-of-Experts Graph Transformers for Interpretable Particle Collision Detection 2025 arXiv preprint arXiv:2501.03432

    2025 arXiv

  20. [28]

    Hardin J and Williams M FastDIRC: a fast Monte Carlo and reconstruction algorithm for DIRC detectors 2016 J. Instrum. 11 P10007 (arXiv:1608.01180)

    2016 arXiv

  21. [29]

    Kalicy G Developing high-performance DIRC detector for the future Electron Ion Collider experiment 2020 Journal of Instrumentation 15 C11006

    2020

  22. [30]

    Holtzman A, Buys J, Du L, Forbes M and Choi Y The curious case of neural text degeneration 2019 (arXiv:1904.09751)

    2019 arXiv

  23. [31]

    Design and performance of the Focusing DIRC detector 2015 Nucl

    Dey B et al. Design and performance of the Focusing DIRC detector 2015 Nucl. Instrum. Methods Phys. Res. A 775 112–131 (arXiv:1410.0075)

    2015 arXiv

  24. [32]

    Science requirements and detector concepts for the electron-ion collider: EIC yellow report 2022 Nuclear Physics A 1026 122447

    Khalek R A et al. Science requirements and detector concepts for the electron-ion collider: EIC yellow report 2022 Nuclear Physics A 1026 122447

    2022

  25. [33]

    ePIC DAQ Working Group 2024 The ePIC Electronics and Data Acquisition Design Status, Fall 2024 https://indico.bnl.gov/event/26584/attachments/62171/106757/ ePIC_DAQ_Design_Status_Oct2024.pdf accessed: July 2025

    2024

  26. [34]

    Wei J, Tay Y, Bommasani R, Raffel C, Zoph B, Borgeaud S, Yogatama D, Bosma M, Zhou D, Metzler D, Chi E H, Hashimoto T, Vinyals O, Liang P, Dean J and Fedus W Emergent Abilities of Large Language Models 2022 Transactions on Machine Learning Research ISSN 2835-8856 https://arxiv...

    2022 arXiv

  27. [35]

    The GlueX DIRC project 2016 J

    Stevens J et al. The GlueX DIRC project 2016 J. Instrum. 11 C07010 (arXiv:1606.05645)

    2016 arXiv

  28. [36]

    Kalicy G The high-performance DIRC for the ePIC detector at the EIC 2024 Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment 169168 Towards Foundation Models for Experimental Readout Systems Combining D...

    2024

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.