Computational Engineering
Papers
Fresh from arXiv, typeset properly. Highlighted words trade on the market.
All of arXiv
Friday 11 September
We study NFT-based reward mechanisms in which a user can create multiple identities and submit fraudulent claims that mature a reward subject to vesting. We assume that the issuer stochastically verifies claims during the vesting period and that identities can be linked into clusters so that the detection of one identity submitting a fraudulent claim causes the whole cluster to be forfeited through a penalty. A farmer's payoff is then non-linear in the number of identities: rewards increase linearly, while the probability of avoiding detection decreases geometrically. We characterise the optimal farming scale in the continuous relaxation of the problem. This allows us to derive a sufficient condition for deterrence and then consider the issuer's choice of reward vesting and claim verification. Vesting reduces the probability that a fraudulent claim is paid but also affects genuine participants, while verification is costly for the issuer. We characterise the sufficient deterrence frontier for the issuer in terms of auditing cost, vesting schedule, and penalty value. When small amounts of audit capacity can be added at negligible marginal cost, vesting alone is not optimal. We also distinguish the role of penalties from that of cluster-level auditing. Finally, we discuss the implications for NFT reward programmes on high-throughput ledgers, such as Hedera.
Number Theory
Iwasawa theory of CM elliptic curves at potentially supersingular primes
We develop a plus and minus Iwasawa theory for CM elliptic curves with potentially supersingular reduction at a prime $p\ge5$. We construct local points that satisfy suitable "jumping trace" relations using the Honda–Demchenko theory applied to height-two formal groups over local fields with small ramification degree. These local points allow us to study plus and minus Selmer groups of CM elliptic curves over the cyclotomic $\mathbb{Z}_p$-extension, to construct the corresponding plus and minus $p$-adic $L$-functions, and to prove an Iwasawa main conjecture relating these objects. As an application, we obtain asymptotic growth formulae for the $p$-primary part of the Tate–Shafarevich groups.
Read the full abstract
Artificial Intelligence
Anchoring Clinical Events in Time: UID-Preserving Multimodal Reconstruction and Source-Grounded Adjudication
Interpreting a patient's hospital course depends on when symptoms developed, tests were performed, and treatments began. Discharge summaries often leave this chronology unclear, while structured electronic health record (EHR) tables timestamp only some of these events. We reconstruct clinical timelines by retrieving structured evidence for narrative events and jointly revising their estimated times. Each extracted occurrence carries a persistent unique identifier (UID) linked to its source span, keeping repeated mentions distinct throughout reconstruction. We also introduce GAVEL, a large language model (LLM) judge that checks disagreements between two timelines against the note and structured record and cites evidence for its verdicts. We evaluated six open-weight models on 40 clinician-annotated discharge summaries. Our pipeline recovered more reference events than LLM-Timeline and TKW2 when each comparison used the same backbone. For GLM-5.2, adding structured evidence improved timestamp agreement with essentially unchanged event recovery; ablations showed that links to timestamped source rows contributed to this improvement. Results varied across the other models, with some timing gains accompanied by fewer recovered events. Both GAVEL judges, GLM-5.2 and MiMo-V2.6-Pro-RL, favored multimodal over text-only GLM-5.2 timelines, while neither clearly distinguished the DeepSeek V3.2 variants. Manual review upheld most sampled GAVEL findings. The source links allow reconstructed times to be checked, and GAVEL identifies disagreements for human review.
Read the full abstract
General Relativity and Quantum Cosmology
Particlelike solutions of the Einstein-Dirac-Higgs equations: ground, excited, and many-fermion states
We present an extended study of the gravitationally localized soliton-like solutions to the minimally-coupled Einstein-Dirac-Higgs equations, embedded in asymptotically Minkowski spacetimes. The equations of motion describing these Yukawa-coupled Dirac stars are generalized to any even number of constituent fermions. We first expand the discussion of two-fermion ground-state solutions initially analyzed in Leith et al. [Phys. Rev. D. 107, 106020 (2023)]. Upon extending our analysis to excited states and many-fermion states, we find that both exhibit different behaviors to their two-fermion ground-state counterparts. In these excited states and many-fermion states, the Higgs field may exhibit stepwise increases and at times decrease within the soliton, which has not been seen for the two-fermion ground states. We also propose the potential cause of a significant degree of ADM-to-fermion mass disparity, which is present in states with strong Yukawa-coupling.
Read the full abstract
Machine Learning
CanvasAnneal: Curriculum Reinforcement Learning for Diffusion Language Models
Diffusion Language Models (DLMs) offer promising parallel generation capabilities but lag behind autoregressive models in complex reasoning and tool-use tasks. While Reinforcement Learning (RL) has recently been applied to enhance DLMs, standard RL approaches suffer from an exploration bottleneck. To address this, we inject reasoning priors from a stronger teacher model to guide RL exploration. In this paper, we introduce CanvasAnneal, a curriculum-guided diffusion RL framework. During the initial RL phase, we warm-start exploration by injecting teacher-generated reasoning traces into the initial diffusion canvas. As training progresses, we gradually remove this guidance and require the model to generate more of the reasoning trajectory independently. Across mathematical reasoning and tool-use benchmarks, CanvasAnneal improves over standard diffu-GRPO on MATH500, Countdown, and Tau2 and substantially accelerates reward improvement on several tasks, while gains are task-dependent. Our results suggest that structured training-time guidance can alleviate exploration bottlenecks in diffusion RL and speed up convergence on harder tasks.
Read the full abstract
Differential Geometry
Prescribed Singular Sets for Z/2-Harmonic 1-Forms on R^n
Z/2 harmonic 1-forms arise naturally as singular limits in gauge theory and calibrated geometry, but the topology that can occur in their singular sets is poorly understood. In this paper we study the flexibility of these singular sets when the ambient Riemannian metric is allowed to vary. We prove that every compact smoothly embedded codimension-two submanifold with trivial normal bundle in a Euclidean space of dimension at least three can be realized as the singular set of a nondegenerate Z/2 harmonic 1-form for a complete metric that is Euclidean outside a compact set. As an application, we use Calabi surgery to construct Z/2-harmonic 1-forms with prescribed local singular sets on closed manifolds of positive first Betti number.
Read the full abstract
Computation and Language
Expert-Space Exploration in MoE Reinforcement Learning
Reinforcement learning (RL) has become central to post-training of large language models. Recent advances in RL for Mixture-of-Experts (MoE) models have primarily focused on improving optimization stability and training efficiency, while treating the expert selection as a fixed component. Since routing determines the sparse computation paths that induce output distributions, expert selection offers an additional source of rollout diversity. Through empirical analysis, we find that perturbing expert routing effectively alters model output and increases rollout diversity, which is similar to increasing the decoding temperature. However, direct perturbation can activate unsuitable experts and substantially degrade rollout quality. Motivated by these observations, we introduce Expert-Space Exploration Reinforcement Learning (ESRL), an architecture-aware framework that explicitly explores the expert-routing space of MoE models. ESRL preserves high-confidence experts as anchors, and restricts stochastic routing to a plausible candidate pool, thereby retaining reliable computation paths. The perturbation strength is further adapted according to router entropy to avoid over-perturbation. To mitigate the routing mismatch introduced by perturbation, ESRL records the expert paths used during rollout and replays them during policy optimization. Experiments demonstrate that ESRL achieves the best performance across MoE backbones with top-K, top-1, and shared-expert routing, as well as across mathematics, science, and code tasks without additional sampling or computational cost. Specifically, ESRL on Qwen3-30B-A3B achieves the best among all compared methods, improving average Pass@1 and Pass@8 over GRPO by 3.2 and 4.5 percentage points, respectively. Further analyses of expert utilization and training dynamics provide insights into how exploiting MoE-specific routing structure benefits RL training.
Read the full abstract
Machine Learning
Benign Loss Landscapes Can Coexist with Worst-Case Hardness
Deep neural networks are expressive enough to contain worst-case targets that can be evaluated in polynomial time but cannot be learned in polynomial time by gradient descent. For practical tasks they nonetheless learn well, raising the question of what non-generic structure of real-world targets enables this. Existing surrogate models cannot pose this question because they either lack hard-to-learn targets entirely (deep linear networks) or cannot evaluate such targets efficiently (kernel methods, infinite-width limits). We study tree tensor networks (TTNs), a model class that generalizes deep linear networks and Tucker decompositions. We show they embed arbitrary read-once Boolean formulas, and thus contain polynomial-size targets that cannot be learned by gradient descent in polynomial time under the same mechanism as neural networks. Despite this, we prove that their loss landscapes are conditionally benign for every realizable target: every local minimum that is minimum-norm is global. Thus, surprisingly, bad local minima are not what distinguishes between typical and worst-case problems in TTNs. Instead, learning difficulty in TTNs can arise from high-order degenerate saddle points, which we show are caused by rank-deficiency. This is explored through a case study of the parity function, illustrating the potential for TTNs to relate landscape geometry to computational hardness.
Read the full abstract
Fluid Dynamics
How turbulent flows grow vorticity at a point
We describe how strong vorticity fluctuations appear in three-dimensional incompressible turbulence driven by a large-scale random force. By combining analytical and numerical methods, we construct a vorticity instanton: the most probable force history driving the flow to a prescribed large vorticity at a given point and time. We show how the smooth, large-scale force produces a long vortex filament with a thin, viscous-scale core. Our main finding is that the standard mechanism of amplifying a vortex tube through radial compression and axial stretching is only effective up to a threshold value. Producing vorticity above the threshold requires an additional ingredient: vorticity waves propagating along the tube. The final vorticity is the result of two colliding waves. This mechanism has striking similarities to the one recently suggested by OpenAI for creating infinite vorticity in finite-time Navier–Stokes blow-up.
Read the full abstract
Quantum Physics
Unified scaling of transmon sub-resonant parametric drive strength limits
Achieving fast gates and high-fidelity readout in superconducting quantum circuits often benefits from the use of large amplitude microwave drives. As the strength increases, unwanted effects such as excitation out of the logical states also become more prevalent. In transmon qubits, coherent controls relying on sub-resonant drives are fundamentally limited by the eventual appearance of strong hybridization in the spectrum. Understanding how this threshold depends on intrinsic device parameters is vital for the design and operation of circuits utilizing the transmon as a source of nonlinearity. In this work, we present an experimental study of 14 transmons with a wide range of anharmonicities, and use comparison with Floquet branch analysis and semiclassical simulations to establish a straightforward relationship between the critical value of the effective drive amplitude $\eta_{\max}$ and the ratio $\gamma=\omega_q/\alpha$, where $\omega_q$ is the transmon frequency and $-\alpha$ is the anharmonicity. For the case where the transmon is used as a parametrically driven coupler, these results further allow us to estimate maximum parametric interaction rates, which informs the circuit design process for optimizing gate rates and fidelities.
Read the full abstract
Differential Geometry
Free boundary minimal surfaces of genus zero
We prove that any embedded genus zero free boundary minimal surface in the ball has first Steklov eigenvalue $1$. This implies the uniqueness of the critical catenoid.
Robotics
Dynin-Robotics: Omnimodal Unified Diffusion Vision-Language-Action Model
Visual goal and dynamics prediction can provide language-conditioned robot policies with both a target outcome and a representation of action-dependent scene changes. We bring these predictions into action generation and selection through a shared trajectory model. Dynin-Robotics implements this formulation on Dynin-Omni, an omnimodal masked-diffusion backbone, representing language, visual observations, goals, and actions as discrete tokens. By varying conditioning and target spans, the same model learns action prediction, action-conditioned next-observation prediction, terminal goal-state prediction, and trajectory-to-instruction reconstruction. These interfaces support test-time scaling through goal prediction, action-candidate evaluation, and joint refinement of action and future-state predictions. We continually pretrain the model on approximately 1.33 million trajectories from 48 Open X-Embodiment datasets and adapt it separately to downstream domains. On two VLABench tasks, robot pretraining improves adaptation within a fixed Stage-2 step budget, and the full objective mixture improves shifted-instruction success over Policy-only post-training under the same coupled decoder. Combining goal guidance with joint action-next-state denoising further improves shifted-instruction success over action-only decoding; the benefit depends on how the predictions are composed. Dynin-Robotics achieves competitive performance on LIBERO and zero-shot LIBERO-Plus, together with a 78.4% average success rate across four manipulation conditions on a Franka Research 3 robot. An optimized block-parallel implementation accelerates model-side action decoding by up to 29.2x relative to the base implementation under the reported profiling setup. These results support shared trajectory modeling as a common interface for learning complementary robot objectives and composing their predictions during control.
Read the full abstract
Differential Geometry
Sectional Curvature Pinching of Two-Step Nilmanifolds
We study the classical problem of sectional curvature pinching in the class of 2-step nilmanifolds, which are necessarily of mixed curvature. We show that the pinching constant of any 2-step nilmanifold lies in the compact interval $[-3, -\frac{3}{2}]$. The upper bound $-\frac{3}{2}$ is achieved by the complex Heisenberg group $\mathrm{Heis}_3(\mathbb{C})$ with a Ricci soliton metric. The upper bound exhibits rigidity: if a simply connected 2-step nilmanifold $N$ has the pinching constant $-\frac{3}{2}$, then $N$ admits a Ricci soliton complex Heisenberg group as a totally geodesic subgroup. On the other hand, the lower bound satisfies non-rigidity: any 2-step nilpotent Lie group admits a metric with pinching constant $-3$. This is derived by showing that there is an open neighborhood $U$ of $\mathrm{Heis}_3(\mathbb{R})\times \mathbb{R}^{n-3}$ in the space of $n$-dimensional 2-step nilmanifolds such that the pinching constant is $-3$ on $U$, and any 2-step nilpotent Lie group $N$ has a metric $g$ such that $(N,g)$ lies in $U$. In fact, if $N$ is not isomorphic to $\mathrm{Heis}_3(\mathbb{R})\times \mathbb{R}^{n-3}$, then there is a curve $g_t$ of metrics on $N$ with $(N,g_t)\in U$, showing that there are uncountably many left-invariant metrics on $N$ such that the pinching constant is $-3$. An algebraic characterization of a 2-step nilpotent Lie group that admits a metric with the pinching constant $-\frac{3}{2}$ is also given, and the pinching constants of various examples are computed.
Read the full abstract
Earth and Planetary Astrophysics
The atomic C/O ratio of KELT-9b
The carbon-to-oxygen (C/O) ratio of a giant planet's atmosphere has long been theorised to hold compositional information that can be traced back to its formation history. Typically, the C/O ratio of an exoplanetary atmosphere is inferred from abundance measurements of major C- and O-bearing molecules such as CO, CO$_2$, H$_2$O, and OH. However, some exoplanets have such elevated temperatures that molecules can be nearly completely thermally dissociated at the pressure levels that observations probe, making it difficult to measure their C/O ratio via these traditional tracers. Here, rather than using molecules, we aim to retrieve the C/O ratio of KELT-9b (T$_\mathrm{eq} = 4000$K) directly from atomic C and O, which are detected in its atmosphere. We analysed two transits of KELT-9b observed with the MAROON-X high-resolution spectrograph, finding absorption cross-correlation signals from atomic C and O as well as refractory metals. From this, we inferred the vertical temperature structure, the relative proportions of volatile and refractory species, and the C/O ratio of KELT-9b using a 1D local thermodynamic equilibrium atmospheric retrieval framework applied only to spectral regions mostly unaffected by non-local thermodynamic equilibrium effects. We measure the atomic C/O ratio of the atmosphere of KELT-9b to be $0.20_{-0.07}^{+0.13}$, which is slightly lower than the stellar value of $0.38\pm0.15$ and significantly below the solar value of $0.59\pm0.07$. We otherwise confirm previous investigations of the terminator region, finding KELT-9b's atmosphere to be thermally inverted and slightly metal-rich. We measure the volatile-to-refractory ratio, a proxy for the ice-to-rock ratio, to be $1.24_{-0.78}^{+1.93}$ $\times$ solar ($[$M$_{\mathrm{vol}}$/M$_{\mathrm{ref}}]$ = $0.09_{-0.20}^{+0.19}$), which is consistent with both KELT-9 and the Sun.
Read the full abstract
Machine Learning
A Unified and Constrained View of Regularization-Based Robust Reinforcement Learning
Regularization-based methods have become a standard approach for training Deep Reinforcement Learning policies against adversarial input perturbations. In this paper, we unify these methods by deriving new upper bounds on the performance gap between the nominal and worst-case policies. Each upper bound is expressed as an existing regularization objective plus a KL-divergence penalty between the nominal and worst-case policies, which further explains why adding a KL penalty improves robustness in practice. Building on these bounds, we formulate robust training as a constrained optimization problem, showing that existing methods correspond to the special case of a fixed Lagrange multiplier. We instead update the multiplier jointly with the policy to automatically tune the regularization weight. Finally, we conduct extensive adversarial evaluations across several continuous control tasks to validate our theoretical analysis.
Read the full abstract
Strongly Correlated Electrons
2D Weak Localization in Trilayer Ruddelsden-Popper Nickelates
Ruddlesden-Popper (RP) nickelates superconduct when pressure suppresses intertwined charge- and spin-density waves. Here, we employ measurements of quantum corrections to magnetoconductivity as a probe of the dimensionality, phase coherence, and scattering mechanisms of the underlying single crystal trilayer RP nickelate Pr4Ni3O10 and La4Ni3O10. We observe signatures of 2D Weak Localization (WL), implying that the in-plane electron transport is in the quantum diffusive regime, while the out-of-plane transport is incoherent, and the electronic ground state is a quasi-2D disordered Fermi liquid. Application of pressure up to 3 GPa continuously suppresses signatures of WL, potentially signaling a 2D/3D dimensional crossover and an eventual pressure-driven transition into a superconductor.
Read the full abstract
Machine Learning
MCRL2: Multi-resource Cross-attention-based Representation Learning-augmented Reinforcement Learning for Cloud Microservice Scheduling
Efficient microservice scheduling is crucial for maintaining load balance across nodes in data centers and ensuring high quality of service. However, achieving this in practice remains challenging due to dynamic resource imbalance under fluctuating workloads, nonlinear coupling across multiple resource dimensions, and the heterogeneity of microservice resource demands. While reinforcement learning-based approaches have shown promise, they struggle to capture the complex interdependencies among heterogeneous resources and neglect the importance of learning informative system representations. To address these limitations, we propose MCRL2, a novel reinforcement learning approach augmented with multi-resource cross-attention-based representation learning for microservice scheduling. Specifically, we first propose MCRL, a novel representation learning approach that captures structured and informative interactions among nodes, resources, and microservices via a multi-resource cross-attention mechanism. Then, MCRL2 augments reinforcement learning through MCRL-enhanced actor-critic architecture combined with a maximum entropy objective, improving system state expressiveness and leading to more stable and effective scheduling decisions. Extensive experiments on real production cluster traces demonstrate that MCRL2 significantly outperforms existing baselines in load balancing, scheduling success rate and average completion time across diverse workload patterns.
Read the full abstract
Artificial Intelligence
Diffusion Models and Concept Formation
Humans organize knowledge into a taxonomy of concepts with nested levels of abstraction and a basic level at which people recognize and name objects with the least cognitive effort. Cobweb is a classic cognitive account of this ability, an incremental learner that builds a probabilistic concept hierarchy by maximizing category utility. We argue that diffusion models, although designed for image synthesis, implicitly perform the same computation. The noisy marginals of a diffusion model are Gaussian smoothings of the data distribution, and the modes of these marginals form a hierarchy that corresponds to a Cobweb tree of probabilistic prototypes in four respects. Both are hierarchical density models, both are hierarchical-Bayesian models with Gaussian prototypes, both treat categorization as score-following that reduces uncertainty, and in both a basic level emerges. We locate this basic level for a diffusion model at an intermediate noise level, where recent analyses show that the reverse process commits to the class identity of a sample. The two models differ mainly in how they represent and learn the taxonomy. Cobweb learns a discrete tree incrementally, whereas a diffusion model encodes a continuous, interpolable hierarchy in a single learned score field fit to the data distribution. We test the correspondence on MNIST and Fashion-MNIST by recovering the diffusion hierarchy through mode-finding and comparing the basic levels of the two models. This reframes diffusion as a cognitive model of concept formation and offers Cobweb a continuous, scalable instantiation.
Read the full abstract
Audio and Speech Processing
Objective Intelligibility Prediction Using Distance Metrics on Speech Foundation Model Representations
High-dimensional representations of pretrained speech foundation models have proven beneficial for objective speech quality and intelligibility prediction. While existing work on neural intelligibility prediction usually leverages such representations for task-specific fine-tuning, in this work we evaluate the usefulness of such representations for intelligibility prediction without any further training. We conduct a layer-wise analysis of multiple speech foundation models, correlating various embedding distances with subjective intelligibility scores. The results show that embeddings extracted from Whisper speech recognition models are best suited, with the last encoder and decoder layers yielding the best correlations when using the Fréchet Audio Distance. Notably, the evaluated distances outperform classical intelligibility metrics and are more robust than Word and Character Error Rates. Further, correlations improve with increasing size of the Whisper model from which embeddings are extracted.
Read the full abstract
Computation and Language
Kraken: LLM-based Speech-to-Speech Translation via Low-bitrate VQ and Dual-path Source Conditioning
Speech-to-speech translation (S2ST) has advanced significantly with speech LLMs, offering the potential for joint optimization and preserving non-linguistic information. However, these models struggle with predicting high-bitrate speech tokens in LLMs, and face the challenge of relying on S2ST training data with ideally aligned speaker identity and prosody. We propose using low-bitrate tokens based on single-layer vector quantization, trained to reconstruct self-supervised learning (SSL) features. We also employ a separate token-to-waveform decoder named Autowave-X, which is also conditioned on the source speech to improve non-linguistic transfer, thereby relaxing the training data constraints. With the integration of these techniques, we propose an S2ST model named Kraken, which augments a pre-trained LLM with speech feature inputs and the low-bitrate token outputs, followed by Autowave-X vocoder. We built the model upon Qwen3-8B and trained it using 150k hours of multilingual and multitask speech data. We demonstrated that our model exhibited better translation quality than SeamlessM4T-Large v2 and Qwen2.5-Omni, along with improved speaker and prosody transfer capabilities.
Read the full abstract
Machine Learning
Robust Policy Optimization via Adversarial Importance Sampling
Significant progress has been made in safeguarding Deep Reinforcement Learning (DRL) policies against input perturbations. Developing robust DRL involves three main stages: algorithm design, implementation, and evaluation. In this work, we identify and address a key limitation at each stage. First, we introduce Adversarial Importance Sampling (Advis), a method that estimates and optimizes worst-case returns directly from trajectories already collected during standard training. Advis satisfies three desirable criteria that prior work does not jointly achieve: it requires no additional environment interactions, no auxiliary networks, and captures long-term robustness. Second, we introduce advrl, a modular PyTorch library that provides clean, single-file implementations of existing robustness methods and adversarial attacks, enabling rapid prototyping and reproducible evaluations. Third, we revisit evaluation under learned adversaries and show that optimal adversarial hyperparameters do not transfer across agents, which can lead to an overestimation of robustness when using a limited set of attacker configurations. Accordingly, we evaluate policies against a large and diverse set of attackers, using 6 to 14 times more configurations than prior work. Finally, experiments on continuous control environments show that Advis achieves stronger worst-case performance and greater training efficiency than existing baselines.
Read the full abstract
Computer Vision
Unified CT and MRI Pancreas Segmentation for Label-Efficient Cross-Modality Subregion Transfer
Robust medical image segmentation across imaging modalities is challenging because of large differences in appearance and intensity distributions. Models trained on a single modality often show substantial performance drops when applied to unseen domains. In this work, we develop a unified 3D pancreas segmentation framework that applies domain-adversarial learning to 4,604 heterogeneous CT and MRI scans to learn anatomical representations. A shared nnU-Net encoder-decoder is trained for whole-pancreas segmentation, with a latent domain discriminator encouraging CT-MRI feature alignment. The learned encoder is subsequently transferred to pancreatic head-body-tail segmentation using limited MRI-only subregion annotations. An average Dice score of 87.31% on the in-distribution test set and Dice scores ranging from 84.20% to 88.09% across external OOD datasets were achieved in whole pancreas segmentation. Dice scores of 80.53% on MRI and 83.05% on CT were achieved for downstream subregion segmentation, without using CT subregion annotations. These results demonstrate that a unified anatomical representation can support both cross-modality pancreas segmentation and label-efficient downstream transfer.
Read the full abstract
Machine Learning
DynSHAP: Towards Explainable Dynamic Survival Analysis
Deep learning models for dynamic survival analysis (DSA) achieve strong predictive performance by incorporating longitudinal patient data, but their black box nature limits clinical trust and adoption. Existing explainability methods cannot handle longitudinal, irregular inputs and functional survival outputs simultaneously, which limits their usability in DSA. We propose DynSHAP, a SHAP framework suited specifically for dynamic survival analysis. It extends common marginal SHAP estimators to this setting by treating time–feature pairs as players in the Shapley game. We further introduce Temporal DynSHAP, which learns linear dependencies in features over time and uses conditional sampling to address them in explanations. When applied to synthetic data with known ground-truth attributions, Temporal DynSHAP recovers temporally dependent features more accurately than marginal estimators for a given state-of-the-art model. Applied to two real-world clinical datasets and two DSA architectures, DynSHAP produces attributions faithful to model learning, allowing medical experts to see which patient information drove the prediction and when.
Read the full abstract
Astrophysics of Galaxies
High-redshift supermassive black hole population from core-collapse in self-interacting dark matter halos
Self-interactions between dark matter (DM) particles facilitate the inter-particle redistribution of energy within the central region of DM halos. Recent studies of dark matter spikes around massive black holes, and the diversity in rotation curves of dark matter rich low-mass galaxies motivate the exploration of self-interaction cross-section $\mathcal{O}(10)\,\rm cm^2/gm$. At such high scattering rates, DM can lead to the formation of supermassive black hole seeds, through the gravothermal collapse of halo cores, under certain cosmological conditions. Quasars near cosmic dawn are powerful probes for examining the formation scenario of high redshift supermassive black holes, and their connection to structure formation. In this work we point out the favorable initial cosmological conditions that are likely to provide the black hole seeds resulting in supermassive black holes, mediated by core-collapse in self-interacting dark matter halos. Employing a semi-analytic prescription of core formation, and realistic mass accretion and halo merger histories, we compare the existing observations of high-redshift supermassive black holes with those predicted for our core-collapse framework. We find the median values of velocity-dependent self-interacting dark matter parameter space $\sigma/m_{\chi}\sim 56\, \rm cm^2/gm$, and $\omega=101\,\rm km/s$, assuming an Eddington accretion rate of unity. Sub-leading values are also presented for sub and super-Eddington accretion rates. We also report the derived self-interacting dark matter model parameters to account for the observed binned supermassive black hole mass functions at high-redshifts.
Read the full abstract
Machine Learning
Quantile-based Loss Filtering for Outlier-Robust Stochastic Gradient Descent
We study loss-based filtering for finite-sum optimization with a subset of corrupted component functions whose gradients may be highly unreliable. Motivated by minimum-loss-based SGD (min-$k$-loss) and quantile-based methods for corrupted linear systems, we propose and analyze a general loss-filtering framework – Quantile-\(k\)-Loss SGD (Q\(k\)L-SGD) – that samples \(k\) component losses at each iteration and updates using an index chosen uniformly from the lower empirical \(q\)-quantile. We prove linear convergence of this family of methods under standard convexity assumptions, requiring the sample size to scale with the number of corruptions and a subset strong-convexity threshold. For the cases when large enough sampling is impossible or undesirable, we give a complementary small-sample probabilistic analysis that covers any sample size $k$ and the convergence behavior depends on the probability of selecting an outlier and on the curvature of the selected good step. Experiments on polynomial regression, regularized logistic regression, and regularized hinge loss show that intermediate quantiles often outperform both standard SGD and min-\(k\)-loss SGD. In particular, min-\(k\) often stalls by repeatedly selecting nearly solved components, while intermediate quantiles retain robustness and produce more informative updates.
Read the full abstract
Machine Learning
Transfer Learning for Evolving Domains
Transfer learning explores how to leverage knowledge from various tasks or domains (sources) to enhance predictive performance in related tasks or domains (targets). Typically, transfer learning research is segmented into several isolated sub-areas (such as domain generalisation, domain adaptation, or multi-domain learning), each making distinct assumptions about target data availability, namely how much data and how many labels are available at training time. However, in many real-world applications, data availability is not fixed but evolves over time, as instances and labels are progressively collected from a new domain. Each of the classical settings then describes only a snapshot of a trajectory that a deployed system must traverse in full. We formalise this trajectory as a transfer learning problem in its own right, Transfer Learning for Evolving Domains (TrED), specified by a data availability process fixed by the environment, a learning protocol that the method is free to choose, and an evaluation criterion that scores the whole trajectory of models rather than a single one. Within this formalism, the classical settings are recovered as regimes that a learner may pass through, rather than as separate problems that TrED concatenates. We then examine the transfer learning literature to identify mechanisms that are promising building blocks for a solution, and find that most methods are tailored to a single regime and that even the strongest existing candidates do not yet optimise the whole trajectory. We argue that TrED is a well-posed and unsolved problem, and an important direction for future research.
Read the full abstract
High Energy Physics (Phenomenology)
Can a minimal radiative seesaw explain the LZ 248 keV event?
We interpret the recently reported 248~keV nuclear recoil event in the LUX-ZEPLIN (LZ) experiment via inelastic dark matter scattering within the minimal Scotogenic model. A sub-MeV mass splitting between neutral inert scalars suppresses low-energy scattering while permitting signals from the high-velocity halo tail. Crucially, co-annihilation with nearly degenerate right-handed fermions accommodates the thermal relic density for dark matter masses up to $\sim {\cal O} (1)$~TeV, extending the viable range significantly beyond the pure inert doublet model limit while evading direct detection bounds. However, to resolve the severe tension with IceCube neutrino limits on solar capture, we extend this minimal framework by introducing a hidden $U(1)_X$ gauge symmetry that naturally leads us to tiny $\lambda_5$ coupling at the one-loop level. This realizes an isospin-violating scenario that suppresses dark matter capture in the Sun while preserving the coherent scattering signal in the Xenon-based LZ detector. We numerically verify that this extended framework naturally generates neutrino masses and satisfies constraints from Big Bang Nucleosynthesis and indirect detection, providing a robust and testable solution to the LZ anomaly.
Read the full abstract
Game Theory
Mitigating Emergent Collusion in LLM Pricing Agents
Recent work shows that LLM-based pricing agents can produce supracompetitive outcomes in repeated oligopoly environments without being explicitly instructed to collude. We reproduce the qualitative prompt-sensitivity effect of Fish et al. using DeepSeek-V3.1: the P1 prompt produces significantly higher prices and profits than P2, although our outcomes are less monopoly-like than the original GPT-4 results. We then evaluate three regulatory interventions: a prompt-only warning, a Harrington-inspired expected-damages payoff regulator, and an active random entrant. The prompt-only regulator reduces but does not eliminate above-Nash pricing. The Harrington regulator brings P1 outcomes close to the duopoly Nash benchmark and removes the statistically significant P1–P2 gap. The active entrant produces the strongest effect, pushing both prompts below the appropriate random-entrant Nash benchmark. Overall, our experiments provide preliminary evidence that interventions that alter incentives or market participation can reduce supracompetitive pricing more effectively than prompt warnings alone.
Read the full abstract
High Energy Physics (Theory)
Thermodynamics of Lorentzian Taub-NUT spacetimes in Einstein-Gauss-Bonnet AdS gravity
We study the thermodynamics of Lorentzian Taub-NUT spacetimes in Einstein-Gauss-Bonnet AdS gravity in arbitrary even dimensions, constructed as $U(1)$ fibrations over Einstein-Kähler base manifolds. We compute the temperature from the surface gravity and the energy as the conserved charge associated with the stationary Killing vector using the off-shell Abbott-Deser-Tekin formalism. Within the same framework, we obtain the entropy as a horizon Noether charge. The result agrees with the Iyer-Wald formula and includes a Gauss-Bonnet correction to the Bekenstein-Hawking area law. Retaining the Misner strings whenever present, we formulate a first law of full cohomogeneity in which the NUT parameter varies independently of the horizon radius. This requires an additional NUT-related thermodynamic variable and its conjugate potential, which we determine explicitly. We also briefly discuss the static limit.
Read the full abstract
Machine Learning
Groupoid-Based Internal State Representations for Reinforcement Learning with Local Symmetries
Symmetries play a central role in reducing the complexity of reinforcement learning problems, yet most existing approaches rely on fixed group actions or predefined state abstractions. Classical reinforcement learning algorithms typically assume a globally structured Markov decision process with uniformly applicable actions and transitions, an assumption that limits their ability to exploit modularity and local, context-dependent regularities present in many realistic environments. We propose a reinforcement learning framework using groupoids to capture local, state-dependent symmetries and support the dy- namic discovery of equivalence structures during interaction. The agent maintains orbit representatives together with transporters that map raw states to canonical forms, enabling learning and decision-making to be performed in a symmetry-reduced space while preserving local distinctions. Empirical results demonstrate that the proposed groupoid-based approach improves sample efficiency and convergence in dense and large-scale environments exhibiting strong partial symmetries, yielding substantial performance gains over standard Q-learning. These findings show that dynamically exploiting local symmetry provides a practical and mathematically principled route to scalable and generalisable reinforcement learning.