Software Engineering
MARGoS: A Configuration-Driven Platform for Reproducible MARL Experiments in ARGoS
Tom Philipp Backert, Mohammad D. Soorati and Javad Ghofrani
Multi-
agent reinforcement learning (MARL) has become a prominent research direction, particularly within collective robotics, where the ARGoS simulator is widely used to study emergent swarm behavior. Yet despite this growth, systematic support for reuse and reproducibility in MARL-based swarm robotics remains limited and insufficiently standardized: a single experiment is assembled from manually aligned simulator files, training scripts, configuration values, paths, seeds, and environment details, so reuse and reproduction stay fragile even within one research group. To address this gap, we introduce MARGoS, a lightweight, configuration-driven platform developed in this project that elevates these experiment-defining inputs to first-class, platform-managed artifacts. MARGoS enables reproducible experimentation through configuration hashing, end-to-end seed propagation, frozen (immutable) configurations, structured logging, environment fingerprinting, and exportable experiment bundles that facilitate sharing, replay, and verification of results across machines and over time. A five-dimension evaluation (architecture, efficiency, reproducibility, usability, and collaboration) supports our claim that managing experiment variation as a platform responsibility makes ARGoS+MARL experiments easier to reuse, reproduce, and hand off. Our code is available at https://github.com/tombackert/margos.
Read the full abstract ▾
High Energy Astrophysical Phenomena
TeV Gamma-Ray Emission in the DA 495 Region from HAWC Observations of J1952+2924, J1954+2837, and J1958+2848
R. Alfaro, C. Alvarez, E. Anita-Rangel et al. (+82 more)
Previous observations of the DA 495 region, which contains the
pulsar wind nebula (PWN) DA 495 and other very-high-energy (VHE; $>100~\textrm{GeV}$) gamma-ray emitters, showed inconsistent results across different VHE experiments, including differences in the measured flux of DA 495 and in the number of reported sources. Our new dedicated analysis using 3071 days of HAWC data resolves these discrepancies. We identify three extended sources, HAWC J1952+2924 (DA 495), HAWC J1954+2837, and HAWC J1958+2848, all with curved spectra, and find a hint of a fourth source. For DA 495, a multi-wavelength analysis, including our results,
radio, X-ray, and other gamma-ray data, reveals that its extended emission is best described by leptonic interactions with an inferred
magnetic field of $B\sim10~\mu\textrm{G}$ and that hadronic components are not required. On the other hand, we interpret the other two sources, which have no confirmed counterparts, as TeV
halos associated with nearby
pulsars, PSR J1954+2836 and PSR J1958+2846, respectively. These results reinforce the interpretation of DA 495 as a leptonic-dominated PWN and add two candidates to the growing population of TeV
halos.
Read the full abstract ▾
All authors R. Alfaro, C. Alvarez, E. Anita-Rangel, M. Araya, J. C. Arteaga-Velázquez, D. Avila Rojas, H. A. Ayala Solares, R. Babu, P. Bangale, A. Bernal, T. Capistrán, A. Carramiñana, S. Casanova, U. Cotti, J. Cotzomi, S. Coutiño de León, E. De la Fuente, N. Di Lalla, R. Diaz Hernandez, M. A. DuVernois, J. C. Díaz-Vélez, C. Espinoza, N. Fraija, S. Fraija, A. Galván-Gámez, J. A. García-González, F. Garfias, N. Ghosh, A. Gonzalez Muñoz, M. M. González, J. A. González, J. A. Goodman, J. Gyeong, J. P. Harding, I. Herzog, D. Huang, F. Hueyotl-Zahuantitla, P. Hüntemeyer, A. Iriarte, S. Kaufmann, D. Kieda, K. Leavitt, W. H. Lee, H. León Vargas, A. L. Longinotti, G. Luis-Raya, K. Malone, O. Martinez, J. Martínez-Castro, J. A. Matthews, P. Miranda-Romagnoli, P. E. Mirón-Enriquez, J. A. Morales-Soto, E. Moreno, M. Mostafá, M. Najafi, L. Nellen, M. U. Nisa, R. Noriega-Papaqui, M. Osorio-Archila, E. Ponce, Y. Pérez Araujo, E. G. Pérez-Pérez, C. D. Rho, D. Rosa-González, M. Roth, A. Sandoval, J. Serna-Franco, M. Shin, A. J. Smith, Y. Son, R. W. Springer, O. Tibolla, I. Torres, R. Torres-Escobedo, L. Villaseñor, X. Wang, Z. Wang, I. J. Watson, H. Wu, S. Yu, S. Yun-Cárcamo, X. Zhang, H. Zhou, C. de León
Logic
Pseudofinite abelian groups and asymptotic classes of finite abelian groups
Carlos Cordoba-Caycedo and Dario Garcia
This paper studies pseudofinite abelian groups and counting properties of definable sets in finite abelian groups and their ultraproducts. First, we characterize pseudofinite abelian groups in terms of Szmielew invariants and, building on results of Macintyre, extend this analysis to obtain a characterization of pseudofinite $\omega$-stable abelian groups. We then give an alternative proof that the class of finite abelian groups forms a multidimensional exact class, explicitly describing the possible cardinalities of definable sets in classes of finite abelian groups. We also investigate $N$-dimensional asymptotic classes of finite abelian groups and provide a series of examples showing that, despite the exactness result, several natural classes of finite abelian groups fail to be $N$-dimensional asymptotic classes for every $N\geq 1$. Finally, in an appendix, we characterize the SU-rank of pseudofinite abelian groups, building on work of Rogers. In particular, we show that for every pseudofinite abelian group $A$, either $\operatorname{Th}(A,+)$ is superstable of finite SU-rank or it is strictly stable.
Read the full abstract ▾
Accelerator Physics
Physics of stabilization and dynamic control of CW magnetrons for superconducting accelerators
G. Kazakevich, R.P. Johnson, I. Gonin et al. (+2 more)
Powerful CW magnetrons were suggested for feeding RF cavities of superconducting accelerators because they provide higher efficiency and lower RF power cost than the traditionally used klystrons, IOTs, or solid-state amplifiers. RF amplifiers driven by a master oscillator serve as coherent RF sources. CW magnetrons are regenerative RF self-oscillators with a huge regenerative gain. This causes regenerative instability with notable noise when a magnetron operates as a self-oscillator, i.e., with the anode voltage above the threshold of self-excitation. An injection locking by a small signal was proposed for phase stabilization of magnetrons. In this case, CW magnetrons with injection-locked (coherent) oscillations generate a notable level of noise. This may preclude the use of CW magnetrons in this mode in superconducting RF (SRF) accelerators. This work discusses our analytical results, PIC simulations, and previously obtained analytical and experimental data regarding the operation and control of CW magnetrons. These results led to the development of methods which are most suitable for SRF accelerators.
Read the full abstract ▾
All authors G. Kazakevich, R.P. Johnson, I. Gonin, V. Yakovlev, Ya. Derbenev
Machine Learning (Statistics)
From Pixels to Privacy-Sensitive Features: An Information-Theoretic Framework for Reconstruction Bounds
Henry Bourne and Robert F. Allison
What can an adversary recover from a model trained with DP-SGD? Its $(\varepsilon,\delta)$ guarantee does not answer this question directly. We derive a distortion bound that can be computed before training and applies even to an adversary who observes the parameter trajectory during training, knows every other training record and the population distribution from which the training records are drawn. The analyst may choose the representation in which error is measured, allowing the bound to target the aspects of reconstruction that are
privacy-sensitive in a given application. Our information-theoretic framework separates two contributions: the information about a record revealed by training, measured through conditional mutual information, and the uncertainty already present in the data distribution, measured through source
entropy. Rate-distortion theory combines them into a lower bound on the expected error of every reconstruction attack. For DP-SGD, the leakage bound depends only on the sampling rates and noise multipliers, while the source
entropy is lower-bounded from a sample with finite-sample confidence. An orthogonal block construction keeps the resulting bound informative for high-dimensional images. We instantiate the framework in pixel space and in LPIPS feature space, obtaining, to our knowledge, the first lower bound on reconstruction distortion measured directly by LPIPS. On a range of image
datasets the bounds establish non-trivial unavoidable expected reconstruction error within heuristically recommended
privacy-budget ranges and beyond. Experiments compare the bounds with strong reconstruction attacks and identify the principal sources of the remaining slack.
Read the full abstract ▾
Machine Learning
DualQuant: Dual-scale Calibration-free Weight-and-Activation Quantization for LLMs
Mahsa Salmani, Sayeh Sharify and Hesham Mostafa
Low-bit inference on accelerators with native 4-bit support requires quantizing both weights and activations to realize the full throughput gains these formats provide. While post-training weight quantization has been extensively studied, activation quantization remains considerably more challenging and typically requires calibration data. In this work, we introduce DUALQUANT, a dual-scale, calibration-free post-training quantization framework that directly minimizes the weight-reconstruction Mean Squared Error (MSE) and, as an immediate by-product, yields per-channel activation rescalings prior to activation quantization. Our key observation is that the scaling factors obtained by minimizing weight-reconstruction error provide effective activation rescalings, enabling competitive activation quantization without calibration data. We formulate DUALQUANT as a non-differentiable weight-reconstruction problem and develop an efficient solver based on variable splitting and block coordinate descent, where each subproblem admits a closed-form update. DUALQUANT provides a unified
algorithm supporting a wide range of formats, including the full MX family and NVFP4. To our knowledge, DUALQUANT is the first dual-scaling method to formally support MX and NV block formats. Across Llama and Qwen models evaluated on WikiText-2, C4, and zero-shot downstream tasks, DUALQUANT matches or exceeds the accuracy of calibration-based SmoothQuant across most models and formats, and substantially outperforms calibration-free baselines, with large gains for MX and NV block formats. We validate these gains with a real packed W4A4 inference path for Llama-3.1-8B on an RTX 4090. The resulting implementation closely matches the accuracy of fake quantization while delivering up to 2.61x
memory reduction and up to 2.46x throughput speedup over BF16.
Read the full abstract ▾
Computer Vision
Shortcut Learning in Geospatial Foundation Model Embeddings for Cross-Disaster Damage Assessment
Ali Shibli, Andrea Nascetti and Yifang Ban
Geospatial foundation models can support disaster damage assessment, but high accuracy on familiar disasters may hide shortcut learning: prediction from event or scene characteristics rather than evidence that transfers to a new disaster. We study
embeddings from TerraMind, Prithvi-EO-2.0, DOFA, and DINOv3 on the multi-disaster
datasets xView2 and BRIGHT. Holding out an entire disaster, instead of randomly splitting image tiles, reduces mean balanced accuracy from 0.807 to 0.638 on xView2 and from 0.818 to 0.568 on BRIGHT. Four sets of tests examine what information supports prediction: event and acquisition context, before/after image inputs, building count, and correct image pairing. Pre-event images remain predictive, and a classifier given only the annotated number of buildings reaches balanced accuracy of 0.629 and 0.653. Combining before/after features performs best on average on xView2, but does not improve over pre-only features on BRIGHT. Shuffling image pairs within the same disaster and damage class causes modest losses. Motivated by these findings, we test whether feature transformations learned only from training disasters improve generalization to unseen disasters. These transformations improve performance, with supervised contrastive learning performing best on xView2 and PCA on BRIGHT. Our results highlight the importance of evaluation on unseen domains for assessing model robustness and show that simple feature-level transformations can improve generalization. Code and experiments can be found at https://github.com/alishibli97/gfm-shortcut-learning.
Read the full abstract ▾
Machine Learning
When Edge Independence Fails: Joint Graph Diffusion with Latent Sociability Priors
Adarsh Jamadandi, Nicolas Keriven and Aline Roumy
Discrete graph
diffusion models corrupt graph structure through independent edge updates with a Markovian noise model. Combined with the exchangeability requirement on the output distribution, the terminal prior is necessarily an Erdős–Rényi (E-R) random graph. This has a fundamental consequence for unconditional graph generation without node features: in the large graph limit, any GNN denoiser applied to an E-R graph approximately outputs an E-R graph, making effective denoising difficult. In practice, generated graphs fail to recover meaningful node heterogeneity and higher-order structure. Previous models introduced specialized node features to compensate, framing this as an expressivity problem, while the role of the prior remained unexamined. We instead address the root cause by introducing the Latent Sociability Prior (LSP) and a join}
diffusion process that co-evolves latent node sociability variables and graph structure, removing edge independence while preserving exchangeability. Unlike existing latent graph
diffusion methods, the sociability mechanism requires no complex encoder: its sole role is to guide edge updates and preserve node heterogeneity throughout the trajectory. We validate our approach on multiple graph generation
datasets and show that the proposed framework more accurately captures graph distributions than baseline methods without node features.
Read the full abstract ▾
Machine Learning (Statistics)
How Well Do Prior-Data Fitted Networks Amortize Bayesian Inference?
Manuel Hülskamp, Julius Kobialka, Emanuel Sommer et al. (+2 more)
Prior-data fitted networks (PFNs) are trained to approximate the
Bayesian posterior predictive distribution (PPD) of a synthetic prior, yet results about the approximation quality are surprisingly thin. Taking an empirical risk minimization perspective, we decompose the deployment error of PFNs into prior, context, and amortization error, and measure the latter against a reference PPD under a realistic prior. Results show that PFNs generally recover the targeted PPD very well. Nevertheless, we find that parts of the residual amortization error persist, especially in rare settings under the prior pretraining distribution, even when scaling the model size, which manifests in disagreement between independently pretrained PFNs and leaves uncertainty about the predictor at inference time. As a remedy, we study multiple types of
Bayesian model averages and find that these can consistently reduce this uncertainty and improve predictions up to full-size TabICLv2 models.
Read the full abstract ▾
All authors Manuel Hülskamp, Julius Kobialka, Emanuel Sommer, Thomas Nagler, David Rügamer
Computer Vision
Neurons++: A Unified Visual-Stream Framework for Image and Video Reconstruction from fMRI with Neural-to-Visual Interpretability
Haonan Wang, Yuchen Wang, Jingyu Lu et al. (+4 more)
Decoding visual experience from fMRI is important for understanding human perception and developing
brain-computer interfaces. Recent generative decoding methods can reconstruct plausible images and
videos, but visual plausibility alone does not ensure that the output is faithfully grounded in
brain activity. In this work, we propose
Neurons++, a visual-stream fMRI reconstruction framework for both static image and dynamic
video reconstruction. Compared with our conference version
Neurons,
Neurons++ extends the paradigm from
video-only reconstruction to visual-stream reconstruction for both images and
videos, and replaces parallel task conditioning with serial cortex-inspired decoding. The model first uses image-fMRI pretraining to obtain a stable spatial-semantic initialization, and then performs subject-specific fine-tuning for image and
video reconstruction separately. The serial design organizes fMRI-derived representations into explicit spatial, semantic,
language-level, and temporal cues. Extensive experiments show that
Neurons++ achieves state-of-the-art
video reconstruction and competitive image reconstruction across standard
benchmarks. Beyond reconstruction fidelity, neural-to-visual analyses indicate that
Neurons++ exhibits qualitative directional controllability under neural perturbation, that its fMRI encoder read-out is sensitive to stimulus-relevant cortical features, and that it supports ROI-guided semantic modulation. Code and model weights are available at: https://github.com/xmed-lab/NeuronsPlusPlus.
Read the full abstract ▾
All authors Haonan Wang, Yuchen Wang, Jingyu Lu, Qixiang Zhang, Lehan Wang, Xuanqi Huang, Xiaomeng Li
Earth and Planetary Astrophysics
The origin of the high heavy-element content of warm Jupiters
Bertram Bitsch
Planet interior models have revealed that warm Jupiters should harbor large heavy element masses (up to 30-40% of the planetary mass). Within the pebble drift and evaporation scenario, inward drifting pebbles can evaporate and enrich the inner disc with vapor to super-
stellar values, accreteable by
giant planets. As H$_2$O ice evaporates in the inner disc, while CO$_2$, CH$_4$ and CO evaporate in the outer disc, the enrichment of the disc with vapor decreases with increasing orbital distance. We use the chemcomp code to model the migration and growth of
planets via pebble and gas
accretion in discs around
stars with different masses and discs with different sizes. We show (i) that pebble drift and evaporation can explain the wide range of derived heavy element contents of
giant planets. (ii)
Giant planets forming around
stars with larger disc sizes can accrete larger quantities of heavy elements, because these discs can maintain a high pebble flux for a longer time. (iii)
Planets forming or migrating interior to the water ice line can accrete larger amounts of heavy elements via the
accretion of vapor enriched gas compared to
planets forming in the outer disc regions. (iv) Metal rich
stars can allow the formation of
planets that are more enriched in heavy elements, if they form in the inner disc regions.
Planets forming in the outer disc regions will be - independently of the disc
metallicity - much less enriched compared. (v)
Planets forming in discs with the same size and total mass, but around
stars with higher mass, are only marginally more enriched compared to their counterparts forming around lower mass
star. Our study suggests that the trend of increasing total heavy element mass with increasing planetary mass is caused by the
accretion of vapor enriched gas in discs with different sizes in the combination with different origins of warm
giant planets from the inner and outer disc.
Read the full abstract ▾
Networking
One-Way Delay Model for 5G Low Latency, High Data Rate Communications
Hamid Hassani, Sonia M. Heemstra de Groot, Ignas G. Niemegeers et al. (+2 more)
Achieving low delay and high data rate communication in 5G requires preventing persistent queue backlogs at path bottlenecks. A key indicator of queue buildup is an increase in one-way delay (OWD) of packets. However, accurately measuring OWD is challenging due to synchronization issues between the sender and receiver. Moreover, using OWD to infer the onset of queueing can be inefficient. For instance, congestion control mechanisms that rely on latency as a congestion signal may result in cellular network underutilization, see, e.g., [1]. Thus, to develop more efficient rate control approaches, it is crucial not only to estimate OWD accurately but also to perform a more fine-grained analysis that provides detailed insights into the queue status. This paper presents a framework for modeling OWD in 5G edge networks, where the wireless link is typically the path bottleneck. %, by modeling it through measurable delay components. We demonstrate that OWD is dominated by queueing delay (QD), particularly at high data rates. We analyze QD by decomposing it into two main components: aggregation delay and channel access time. Aggregation delay can be accurately measured at the user equipment (UE), and channel access time can be estimated based on network conditions. Our key contribution shows that QD, while challenging to measure directly at the base station (BS), can be closely approximated by combining these components. We validate this model through NS3 simulations under varying network loads and single- and multi-UE scenarios. This framework provides a foundation for developing advanced mechanisms for congestion control and bottleneck detection in 5G networks.
Read the full abstract ▾
All authors Hamid Hassani, Sonia M. Heemstra de Groot, Ignas G. Niemegeers, Kishor C. Joshi, George Exarchakos
Audio and Speech Processing
OpenSVBench: Diagnosing Speaker Verification Generalization Across Deployment Scenarios
Qituan Shangguan, Bosen Xu and Shuai Wang
Speaker verification (SV) systems now achieve very low error rates on standard public protocols, but those results do not fully describe deployment generalization. VoxCeleb-style evaluations remain useful anchors for English in-the-wild media
speech. Scenario-specific evaluations and challenges are also valuable. However, neither setting directly shows whether the same system remains reliable across diverse deployment scenarios. We introduce OpenSVBench, a public
benchmark for cross-scenario SV evaluation. Its fixed public core contains 26 trial sets organized into 11 scenario groups, with 9.15M verification trials, frozen target/non-target keys, and metadata for scenario-wise reporting. Around this core, the online leaderboard acts as a diagnostic evaluation interface rather than only a ranked list. It accepts score packages from new systems, evaluates them with frozen keys and common scoring, and reports trial-set breakdowns, scenario-group metrics, macro EER/minDCF, model-wise profiles, and scenario-wise comparisons. We evaluate 18 public checkpoints spanning x-vector, ECAPA-TDNN, CAMPPlus, ResNet and ERes variants, ReDimNet2, SAM-ResNet, and pretrained
Transformer speaker recipes. The results show clear cross-scenario gaps: systems with similar macro EER can have different scenario profiles, strong performance in one scenario does not guarantee strong performance in another, and same-scenario trial sets can still expose source and protocol gaps.
Read the full abstract ▾
Game Theory
Asymmetric Information in Dynamic Lending Games
Kristijan Atanasov, Frederik Mallmann-Trenn and Carmine Ventre
We study a dynamic lending game with limited but reusable supply. Borrower arrivals and loan durations are
stochastic. Demand is allocated cheapest-first, subject to available supply, so a lender's quote affects both its current payoff and its ability to serve future borrowers. Motivated by industry practice, we study this problem through the lens of information disclosure: Lenders observe either only their own available supply or also their rivals' aggregate available supply. We ask how providing this information to all, some or none of the lenders affects payoffs as demand varies relative to supply, and how these effects change when a fixed total supply is divided among different numbers of lenders. Interestingly, in the case of asymmetric information disclosure, total payoffs can increase, while total lending remains the same. We identify sufficient mathematical regimes in which this phenomenon occurs. Who receives supply information matters differently as demand rises. For the case of two lenders, informing only the larger lender yields higher total lender payoff at lower demand, while informing only the smaller lender can yield the highest total lender payoff at higher demand. For a platform deciding which lenders can observe rival supply, the results suggest that disclosure can change the fees paid by borrowers without changing how much is lent. Wider access to supply information need not lower those fees: its effect depends on who receives the information and the balance between demand and supply.
Read the full abstract ▾
Astrophysics of Galaxies
A dust-reddened quasar in a [C II]-luminous host at z=6.55
Daming Yang, Joseph F. Hennawi, Rychard Bouwens et al. (+18 more)
Most known
quasars at $z>6$ were selected through their rest-frame UV emission. Only a small number of reddened
quasars have been identified at these
redshifts, leaving their
nuclear and host-
galaxy properties poorly constrained. We present JWST/NIRSpec spectroscopy and ALMA band 6 observations of the
quasar J1440+0019 at $z=6.5488$, revealing its
dust-reddened nature and jointly characterizing its
nucleus and host
galaxy. The continuum is best described by an SMC-like extinction law with $E(B-V)=0.20\pm0.03$. Three absorption systems detected in N V, Si IV, and C IV are blueshifted by approximately $900$–$3600$ km s$^{-1}$ and likely trace an ionized
nuclear outflow. The broad Mg II line and extinction-corrected continuum yield a black-hole mass of $\log(M_{\rm BH}/M_\odot)=9.04\pm0.42$, a bolometric luminosity of $L_{\rm bol}\simeq1.1\times10^{47}$ erg s$^{-1}$, and an Eddington ratio of $\lambda_{\rm Edd}\simeq0.8$. ALMA observations further reveal its host
galaxy to be among the most [C II]-luminous known at $z>6$, with $L_{\mathrm{[C II]}}=(7.21\pm0.25)\times10^9 L_\odot$. For a fiducial
dust temperature of 47 K and emissivity index $\beta=1.6$, the
dust continuum implies $L_{\rm TIR}=8.9\times10^{12} L_\odot$ and $M_{\rm dust}=4.0\times10^8 M_\odot$. The inferred dynamical mass places J1440+0019 in the region of the $M_{\rm BH}$–$M_{\rm dyn}$ plane occupied by high-
redshift blue
quasars. These observations reveal a moderately reddened, rapidly accreting
quasar in a luminous,
dust-rich host during reionization. Its measured
nuclear and host-
galaxy properties provide an observational reference for identifying and characterizing similar systems in future systematic searches with wide-area infrared surveys.
Read the full abstract ▾
All authors Daming Yang, Joseph F. Hennawi, Rychard Bouwens, Timo Kist, Eduardo Bañados, Hiddo S. B. Algera, Aaron J. Barth, Sarah E. I. Bosman, Frederick B. Davies, Anna-Christina Eilers, Xiaohui Fan, Victoria A. Fawcett, Yuming Fu, Anniek J. Gloudemans, Francesco Guarneri, Bin Jia, Elia Pizzati, Huub Röttgering, Jan-Torge Schindler, Feige Wang, Jinyi Yang
Computer Vision
Gaussian Splatting with Reflections Guided by What is Seen
Yiming Liang, Tianyu Xiao and Hiroshi Ishikawa
While novel view synthesis methods like Neural Radiance Fields (NeRF) and 3D Gaussian Splatting (3DGS) have shown impressive performance on common scenes, they often struggle with capturing view-dependent specular effects, particularly near-field reflections. Existing methods utilize environment maps, which only represent lights from infinity, or environment Gaussians, which rely solely on indirect supervision from reflections, are insufficient for accurately modeling such effects. Here, we propose a framework based on 2D Gaussian Splatting (2DGS) that leverages directly observable environment information in input images. Our approach also integrates a hybrid reflection appearance representation and view-dependent specular modeling. Experiments on public
benchmarks and a newly curated near-field reflection
dataset demonstrate that our method achieves superior rendering quality and
geometric accuracy.
Read the full abstract ▾
Databases
ProvQML: Efficient Provenance Capture for Machine Learning as Recursive Aggregate SQL [Extended Version]
Nkechi J. Akinwale and Seokki Lee
Expressing ML
algorithms as recursive aggregate SQL queries over distributed systems enables large-scale declarative data analytics, yet no existing system supports scalable provenance capture, necessary for ML explainability. We present ProvQML, a provenance-aware recursive aggregate SQL engine that efficiently captures fine-grained provenance while evaluating ML
algorithms as recursive aggregate queries. We capture two types of provenance: (i) local provenance that records the contribution of each training instance to gradient computation at a specific model state, and (ii) global provenance that encodes the sequence of model parameter updates across the iterations. Extensive experiments on widely used ML
datasets demonstrate that our approach efficiently captures both provenance for various ML models over widely used ML
datasets.
Read the full abstract ▾
Machine Learning
Matched-Context Group Policy Optimization for LLM Agent Training
Chenxu Zhang, Yaorui Shi, Wenlong Zhang et al. (+3 more)
Training long-horizon
language model
agents requires assigning credit to intermediate decisions from sparse terminal feedback. Within a finite group of sibling rollouts, useful action comparisons depend on their context, evidence for both alternatives, and how that evidence enters learning. We propose Matched-Context Group Policy Optimization (MCGPO), a procedure for constructing weighted action preferences from already collected trajectories. MCGPO defines comparison contexts with pre-action state summaries and observed rubric progress, annotated by task rules. Two-action support selects among ordered exact and coarse matching rules. Fixed folds separate candidate decisions from the terminal outcomes used to orient their preferences, and support-dependent evidence weights scale an auxiliary pairwise objective alongside shared GraphGPO task learning. MCGPO reuses rollout data and each response's original history, preserving critic-free training. Across six environment-backbone settings with three training seeds each, MCGPO improves mean success over matched GraphGPO: by 0.47-1.46 percentage points on ALFWorld and WebShop with Qwen2.5-1.5B and 7B, and by 3.96 and 5.16 points on WebShop with Qwen3-4B and 8B. The ALFWorld gains build on baseline success above 92%, where remaining headroom is limited. Integrated success improves in three of four Qwen2.5 settings. Three-seed factorial controls support opposite-fold evidence and evidence weighting; joint state-rubric matching achieves the highest mean in all four representation-control settings.
Read the full abstract ▾
All authors Chenxu Zhang, Yaorui Shi, Wenlong Zhang, Jianchao Tan, Zihan Wang, Yiming Cheng
High Energy Physics (Phenomenology)
Utilizing Neural Networks for Solving Functional Renormalization Group Flow Equations
J. Gurzó and G. Fejős
A novel physics-informed neural networks (PINNs) based approach for solving functional renormalization group flow equations is presented. We investigate the finite temperature phase transition in $O(N)$ models without relying on the large-$N$ approximation, which was used in earlier attempts with similar methods. We argue that these previous solutions cannot easily be generalized for more complex
symmetries, and hence our approach can provide new insight. Our technique relies on a series of ansätze that combines the small- and large-field approximation schemes with a neural network, leading to a highly accurate approximation over the entire field range. We compare the PINNs approach with traditional grid-based Runge-Kutta results and show that the network can successfully learn the temperature, field, and scale dependencies of the effective potential in the local potential approximation.
Read the full abstract ▾
Analysis of PDEs
Scattering for the KP-II equation
Xi Chen
We prove scattering in $L^2(\mathbb{R}^2)$ for real-valued solutions of the Kadomtsev–Petviashvili II (KP-II) equation with localized initial data. No smallness assumption is made, no zero mean condition is imposed, and the admissible data include all Schwartz functions. The proof uses the inverse scattering transform attached to the heat operator. It rests on three ingredients: an a priori bound for the scalar $\bar\partial$ problem which depends on the size of the scattering coefficient but not on its oscillating phase, a nonlinear Plancherel identity proved by analytic continuation from small data, and an estimate showing that the nonlinear part of the reconstruction is negligible on the propagation cones of free waves. To the best of our knowledge, this is the first result establishing scattering for the KP-II equation for every real-valued Schwartz initial datum.
Read the full abstract ▾
Analysis of PDEs
Asymmetric minimal mass blow-up for the quintic NLS on a star graph
Mattia Buratti, François Genoud and Vladimir Georgiev
This work is concerned with the focusing quintic nonlinear Schrödinger equation on a
star graph with two edges, with a delta interaction of arbitrary strength at the vertex. We construct a non-radial, minimal mass, finite time blow-up solution, whose singularity forms at an arbitrarily prescribed interior point of one edge. Our approach is based on a suitable cut-off of the standard pseudoconformal blow-up solution of the focusing quintic NLS on the line, inspired by Ogawa-Tsutsumi and Burq-Gérard-Tzvetkov. A noteworthy feature of our construction is the persistence of the domain of the
Hamiltonian under the flow, obtained by the persistence of regularity method of Kato. We also identify a two-parameter family of distinct blow-up profiles, which should play a central role in the classification of non-radial minimal mass blow-up solutions on the
star graph. Our construction is generic and can be extended to a
star graph with an arbitrary number of edges. It can certainly also be adapted to other types of metric graphs.
Read the full abstract ▾
Optimization and Control
Estimation, Sensitivity Analysis, and Optimization of Distortion Risk Measures via Simulation
Hao Cao, Jian-Qiang Hu and Jiaqiao Hu
We present a simulation-based method for the estimation, sensitivity analysis, and optimization of distortion risk measures (DRMs), a class of performance measures that capture a wide range of risk attitudes through an underlying distortion function. Existing simulation approaches often approximate DRMs using numerical partitions or empirical distributions. We show that, when the distortion function is piecewise
polynomial, the nonlinear structure of a DRM can be handled exactly, without introducing distortion discretization error, using only simulated system outputs. This representation enables the development of a computational method that integrates finite-difference techniques for sensitivity analysis and for solving a broad class of differentiable risk-based optimization problems. Under suitable conditions, we establish strong consistency and convergence rates for DRM estimation, sensitivity analysis, and optimization. We further extend the approach to non-
polynomial distortion functions through
polynomial approximation. Numerical experiments on synthetic and calibrated financial examples demonstrate the effectiveness and robustness of the proposed approach across a range of applications.
Read the full abstract ▾
Machine Learning
Studying the Effects of Task Diversity Using Small Mathematical Worlds
Emily Xia Meng, Xuanyu Yang, Junaid Hasan et al. (+2 more)
Basic mathematical tasks like modular addition have been a valuable tool for illuminating aspects of neural network behavior that are challenging to see in noisy, unverifiable, large-scale, real-world settings. These studies have typically focused on a single task. However, modern
language models are exposed to a vast number of tasks despite being trained with a next-
token prediction objective. In this work, we introduce two mathematical task ecosystems, IntWorld and PermWorld, built from distinct but related tasks on integers and permutations. Both are designed to simulate task diversity within a cohesive mathematical 'world' where tasks are distinct but related. Because the
datasets are procedurally generated, we can precisely vary the number of tasks, allowing us to begin to disentangle the role of task variety from other factors in training. After describing the
datasets, we provide some initial results obtained from these
datasets which support the idea that increasing task diversity increases the richness of a model's internal representations.
Read the full abstract ▾
All authors Emily Xia Meng, Xuanyu Yang, Junaid Hasan, Michael Ruofan Zeng, Henry Kvinge
Networking
RAPID: Proactive Rate Control for Enabling High Data Rate and Low Latency in 5G Edge Networks
Hamid Hassani, Sonia M. Heemstra de Groot, Ignas G. Niemegeers et al. (+2 more)
Many 6G use cases such as tactile Internet and eXtended reality applications would simultaneously require a high data rate and low latency wireless connectivity.It is increasingly being realized that the current transport protocols relying on packet loss and delay are insufficient to satisfy the data rate and latency requirements of such applications. To address this problem, we propose RAPID, a transport layer solution designed for private 5G edge networks, where the bottleneck is located at the base station. We show that by regulating the send rate to maintain a target physical resource block utilization, we can achieve high data rate, low latency communication. We design a simple yet effective policy that adapts to the dynamic nature of wireless channel conditions. In a multi-UE scenario with two RAPID flows and one TCP Cubic flow, our testbed results show that RAPID flows achieve significant improvements in one-way delay, i.e., 72x and 55x lower compared to TCP Cubic, with negligible packet loss at the cost of modest goodput reductions of 16% and 27%, respectively.
Read the full abstract ▾
All authors Hamid Hassani, Sonia M. Heemstra de Groot, Ignas G. Niemegeers, Kishor C. Joshi, George Exarchakos
Soft Condensed Matter
Arrested phase separation and amorphous patterns in charged binary fluids
Logan H. Grant, Ana Bensabat, Eva Pospivsilova et al. (+6 more)
Frustration generated by competing interactions can stabilise patterns with tunable lengthscales which evade indefinite coarsening and macroscopic phase separation. Here we investigate a charged binary fluid in which the tendency toward phase separation is opposed by screened electrostatic repulsion. Combining analytics and numerical simulations, we identify a regime of arrested phase separation characterised by finite-wavelength instability. The resulting states include droplet and labyrinthine phases in two dimensions, and droplets, concentric shells and tubular structures in three dimensions. Surprisingly, some of these structures are amorphous morphologies that emerge reproducibly from different initial conditions. Free-energy estimates suggest that these disordered structures may be thermodynamically stabilised by frustration, rather than being kinetically trapped. Our results identify a simple mechanism through which the competition between phase separation and screened electrostatics can generate
equilibrium amorphous microphases with glassy features.
Read the full abstract ▾
All authors Logan H. Grant, Ana Bensabat, Eva Pospivsilova, Joost de Graaf, Alexander Morozov, Giuseppe Negro, Ciprian G. Pruteanu, Tyler N. Shendruk, Davide Marenduzzo
Analysis of PDEs
Travel time tomography for elliptically anisotropic elastic planar media
Joonas Ilmavirta, Teemu Saksala and Hjørdis Schlüter
We study the gauge of the travel time
tomography problem for elliptically anisotropic planar media. Under this assumption the respective slowness surfaces consist of an inner ellipse and an outer circle. Hence, the singularities of seismic waves propagate on geodesics given by two Riemannian metrics of which the one related to shear waves is conformally Euclidean. We show that if both of these metrics are lens rigid then, up to boundary fixing diffeomorphisms, each elliptically anisotropic stiffness tensor is uniquely determined by the lens data whenever it has at least one isotropic point. If none of those points exists, then there are exactly two elliptically anisotropic stiffness tensors with the same lens data. Clearly, not all boundary fixing diffeomorphisms preserve the imposed structure of the slowness surface. We show that the set of well-behaved diffeomorphisms is never a subgroup of all boundary fixing diffeomorphisms. Our proofs are based on an interplay of
algebra and geometry, and the key of the proofs is a characterization of the
elliptic anisotropy via several
polynomial conditions.
Read the full abstract ▾
Differential Geometry
Topping's conjecture for solutions of harmonic map flow with isolated bubbles
Melanie Rupflin, Alex Waldron and Sebastian Woodward
By establishing a weighted Lojasiewicz-Simon inequality for maps which are close to a fixed harmonic map modulo scaling, we prove Topping's
conjecture on continuity of the body map in the case of finite time singularities at which only isolated bubbles form.
Rings and Algebras
The $E_6$ normal forms: semisimple deformations, geometry, DT invariants, and the $E_{6,\infty}$ order
Shengyong Pan
We study the completed Jacobi
algebras of $x^3+xy^3+p(y)$ for $\operatorname{ord}_yp=n\ge4$. The explicit normal-form morsification, the filtered length and centre proofs, and the cyclic terminal identities give finite free semisimple deformations. For every parameter of finite order we construct a
polynomial curve neighbourhood and a local projective small contraction; the resulting
geometric GV and Donaldson–Thomas calculations identify the block multiplicities. The one-dimensional limit $E_{6,\infty}$ is studied separately as an order: we determine its normalization, conductor, radical-idealizer chain, and maximal overorders. The canonical chain yields a nine-dimensional finite quotient whose Ringel dual is isomorphic to its opposite. An expanded construction gives a choice-independent $22$-dimensional quotient and a $40$-dimensional Ringel dual; a corner quotient relates the two finite constructions. The length and centre formulas are proved by a filtered comparison with a homogeneous
matrix order and by terminal identities at the exceptional endpoint.
Read the full abstract ▾
Artificial Intelligence
A Privileged Teacher That Only Scores Cannot Search: Critique-Conditioned On-Policy Self-Distillation for Reasoning
Mayug Maniparambil, Ayush Rai, Ekansh Verma et al. (+2 more)
On-policy self-
distillation promises to replace the sparse outcome signal of verifiable-reward RL with dense
token-level credit: a teacher conditioned on the reference solution scores the student's own rollout, and the resulting distributional shift is distilled back into the policy. We investigate whether this scoring signal actually contains corrective credit. Using gold error spans in student rollouts, we test the two signals consumed by self-
distillation objectives, the full-distribution KL and the
token log-ratio, for two properties: whether they localize errors and whether they push probability away from erroneous student
tokens. They exhibit neither of these properties nor can they reliably distinguish correct from incorrect rollouts within the same problem. We argue that this failure is structural: locating an error in a
reasoning trace is itself a search problem, while a scoring teacher is never explicitly asked to perform that search. The same model, with the same privileged information, localizes errors substantially better when allowed to generate a critique. This forms the motivation for critique-conditioned self-
distillation: first generate a critique of the current rollout, then condition the scoring teacher on that critique. The resulting signal localizes errors, is corrective within error spans, and shifts probability toward
tokens that initiate backtracking such as “Wait”, “But”, and “Actually”. We further show that the critique is useful both at test time, where resuming a wrong rollout with the critique in context converts math problems the model cannot solve by sampling, and in cold-start GRPO, where it provides an advantage signal on
prompts for which verifiable rewards alone provide none.
Read the full abstract ▾
All authors Mayug Maniparambil, Ayush Rai, Ekansh Verma, Noel E. O'Connor, Fergal Reid
Mathematical Physics
Perturbed isotropic oscillators with inversion symmetry
Antonella Marchesiello, Rita Mastroianni and Giuseppe Pucacco
We consider the normal form of a 1:1 detuned resonance that can be related, via a double averaging process, to the three body problem of celestial mechanics. In view of the
symmetry in the gravitational coupling between the
planets, the obtained normal form exhibits inversion
symmetry. We provide an analytical prediction of the possible bifurcations in terms of the detunings and the other physical parameters of the system. We show that the bifurcation scenario is rich. It includes bifurcation sequences where the unstable orbit appearing after a saddle-node bifurcation can have a reversed pitchfork bifurcation, where it turns into stability with the birth of two unstable orbits, which eventually reconnect with one of the two quasi-normal modes. Furthermore, another scenario shows a standard pitchfork in which one of the quasi-normal modes becomes unstable, giving rise to a pair of stable periodic orbits. By introducing coupling and asymmetry parameters, we provide an effective way to discern all the possible sequences of bifurcations.
Read the full abstract ▾