Computation and Language
Exact Feedback Is Not Control: Evaluating Text-based Closed-Loop Revision in LLMs
Haitong Jiang, Chunlin Liu, Yile Wang et al. (+1 more)
Closed-loop revision is increasingly used in large
language model (LLM) applications, but failures may reflect incomplete feedback or ineffective responses to correct feedback. We introduce a fixed-budget revision protocol with deterministic verifiers that report all remaining violations across exact-length, lexical, and compositional constraints. Fixing feedback correctness and completeness isolates model-side revision behavior. Across 19 open- and closed-source models, controller-level mean final joint success ranges from 17.4% to 99.8%, with substantial cross-model gaps persisting under identical initial drafts. Controlled experiments reveal reproducible model-specific responses to exact feedback. Post-training and scale reshape these responses without consistently bringing them closer to exact correction. Across all constraint families, failed trajectories often repeat earlier outputs, and prior recurrence is associated with lower subsequent recoverability. Matched-state interventions show that removing earlier dialogue while holding the current draft and feedback fixed changes recurrence escape without reliably improving final success; effects depend on the model, task, and trigger-state composition. Exact feedback makes revision errors observable, but does not make the closed loop reliable. Code and reproduction instructions: https://github.com/kevinjiang0121-cyber/exact-feedback-code.
Read the full abstract ▾
All authors Haitong Jiang, Chunlin Liu, Yile Wang, Yuhong Feng
Systems and Control
EvEMTBench: An Open Benchmark for Machine Learning in Power System Protection
Julian Oelhaf, Georg Kordowich, Christian Bergler et al. (+3 more)
Studies of machine-learning-based power system protection are difficult to compare because task definitions, measurement access, data partitions, metrics, and generalization conditions often differ. EvEMTBench addresses this gap with an open, executable, and versioned
benchmark that fixes these evaluation choices while leaving model design open. Across four grids spanning 20-345 kV, it defines 12 protection and event-analysis functions instantiated as 24 scored tasks and supports structured evaluation across observability conditions, predefined distribution shifts, and zero-shot and fine-tuned cross-grid transfer. Committed partitions, leakage controls, and reproducible reporting provide a common basis for comparing future methods. A reference evaluation spanning trivial, conventional, feature-based, and deep-learning baselines shows that wider observability is not uniformly beneficial, shifted conditions can reveal failures not apparent in-distribution, and cross-grid transfer is substantially stronger for fault detection than for fault localization. Protection-relevant diagnostics identify failure modes not apparent from primary metrics alone. EvEMTBench therefore makes generalization in machine-learning-based protection an explicit and reproducible evaluation problem.
Read the full abstract ▾
All authors Julian Oelhaf, Georg Kordowich, Christian Bergler, Andreas Maier, Johann Jäger, Siming Bayer
Statistical Mechanics
Universal splitting of nonequilibrium phase transitions in driven Potts heat engines
Vitória T. Henkes, Gustavo A. L. Forão, Andre C. Barato et al. (+1 more)
We investigate nonequilibrium phase transitions and thermodynamic properties of driven Potts models coupled to two thermal reservoirs at different temperatures. The interplay with the multi-state structure of the Potts model leads to distinct phase-transition scenarios. The proposed driving scheme breaks the
symmetry between the Potts states and gives rise to multiple transitions within the ordered phases, where the phases are characterized by different numbers of stable fixed points. This number can vary from $q$ fixed points to $1$ fixed point, corresponding to $q-1$ distinct transitions. The driving scheme strongly affects the thermodynamic operation regimes, allowing the system to operate as a heat engine. We demonstrate that these phenomena persist for both a mean-field (MF) model and a two-dimensional model, confirming the robustness of the results with respect to dimensionality. For the MF case, we also develop a phenomenological description in the strongly ordered regime that yields analytical expressions for the power, heat currents, and efficiency. Our results show how multi-state collective interactions, nonequilibrium driving, and thermal bias jointly generate a rich phase structure and the emergence of a collective heat engine.
Read the full abstract ▾
All authors Vitória T. Henkes, Gustavo A. L. Forão, Andre C. Barato, Carlos E. Fiore
Quantum Physics
Yan-Bo Jiang, Xiao-Hui Wang, Kun Zhang et al. (+1 more)
Grover's
algorithm is optimal in terms of oracle queries. We can trade accuracy for speed, which gives rise to the
quantum partial-search
algorithm. The partial-search
algorithm is implemented using two kinds of Grover operators, global and local, where the former is the standard Grover operator for full search and the latter has a
diffusion operator acting on the search subspace. The global-local-global sequence, also known as the Grover-Radhakrishnan-Korepin (GRK)
algorithm, has been proved optimal in the oracle-query metric. In this work, we show that the alternating sequence of global and local Grover operators, formed by repeatedly applying a fixed product of global and local Grover operators, can achieve a lower expected
circuit depth than the GRK
algorithm. Through systematic analysis, we obtain its exact success probability, expected depth, and asymptotically optimal parameters. We derive the boundary, characterized by the ratio between the depths of the oracle and the global
diffusion operator, separating the depth-optimal and oracle-optimal partial-search
algorithms. When the depths of the oracle and the global
diffusion operator are comparable, our proposed alternating partial-search sequence can reduce the minimum expected
circuit depth by more than 20% compared with the GRK
algorithm.
Read the full abstract ▾
All authors Yan-Bo Jiang, Xiao-Hui Wang, Kun Zhang, Vladimir Korepin
General Relativity and Quantum Cosmology
A Catalog-Wide Study of Gravitational-Wave Residuals in GWTC-4 and GWTC-5
Junlin Qin, Zhan-Feng Mai, Dicong Liang et al. (+2 more)
Improved sensitivity of the LIGO-Virgo-KAGRA detectors has enabled the accumulation of a large gravitational-wave event catalog, providing an opportunity for catalog-wide residual analyses to assess waveform-model performance. We study GWTC-4 and GWTC-5 events with signal-to-noise ratio $>12$ in at least one detector, applying three goodness-of-fit tests to the normalized Q-transform energies of whitened residuals obtained by subtracting the IMRPhenomXPHM and SEOBNRv5PHM waveform models. Across the catalog, the residuals are consistent with Gaussian noise, with no evidence of systematic deviations in either multi-detector events or valid single-detector events. For events exhibiting strong model dependence in parameter estimation, variations between waveform models likewise do not produce persistent residual structures. These results demonstrate that parameter-estimation biases associated with waveform modeling do not necessarily manifest as detectable features in the residuals. Residual analyses therefore provide a complementary diagnostic in data space for validating gravitational-wave waveform models within the detector sensitivity.
Read the full abstract ▾
All authors Junlin Qin, Zhan-Feng Mai, Dicong Liang, Hai-Tian Wang, En-Wei Liang
Machine Learning
RL Starts before RL: On Policy Distillation for Better Reinforcement Learning
Shuai Dong, Yongfu Zhu, Yuqi Xu et al. (+26 more)
Reinforcement learning (RL) improves
reasoning, but its performance depends on the policy from which training begins. We study on-policy
distillation (OPD) as a preparation stage for RL and ask whether its benefits extend beyond improvements in the distilled model's initial accuracy. Under shared RL settings, students initialized with OPD reach higher final performance than those trained with direct RL or supervised fine-tuning followed by RL. This advantage can emerge even when OPD produces little immediate improvement in accuracy. Pre-RL Pass@k does not fully explain the benefit: similar or even higher values do not necessarily lead to better performance after RL. Behavioral analyses point to
alignment with the teacher's distribution beyond top-1 agreement as a possible explanation. Such
alignment may favor higher-quality
reasoning paths while retaining alternatives that RL can further refine using outcome feedback. We further examine how trajectory sources and divergence objectives affect the value of
distillation for subsequent RL. Standard reverse-KL OPD performs better before RL, but forward-KL OPD overtakes it afterward. Student rollouts outperform teacher rollouts under both objectives before and after RL. These findings highlight the importance of both objective choice and the states receiving supervision for subsequent RL. Our results support OPD as preparation for RL and favor forward KL when OPD is followed by RL in our comparison.
Read the full abstract ▾
All authors Shuai Dong, Yongfu Zhu, Yuqi Xu, Weichu Xie, Liuwenpu, Ziyue Wang, Kaiwen Tuo, Congcong Wang, Siyuan Wang, Wenqi Shao, Shuai Yang, Ji Zhao, Caoyuan Ma, Wenzheng Chang, Taiqiang Wu, Xinlei Yu, Hongrui Wu, Xiaoxuan He, Fangke Chen, Dianyi Wang, Kanghui Tian, Sirry Chen, Xingyu Liu, Xiangnan Wu, Jiawei Guo, Haowen Hou, LingHan Chen, Zhongyu Wei, Jiaqi Wang
High Energy Physics (Experiment)
Integer Quantization of Graph Neural Networks for Real-Time FPGA Track Finding
Andrea Cardini, Pelayo Leguina, Elena Aller et al. (+1 more)
Real-time track finding for displaced-
muon signatures in the CMS Level-1 trigger must operate under strict fixed-latency constraints of 12.5 $\mu$s while processing high-throughput detector data. Because
muon hits map naturally onto sparse, irregular graphs, graph neural networks (GNNs) are attractive candidates; however, mapping message-passing models to field-programmable gate arrays (FPGAs) requires careful co-design of numerical precision, microarchitecture, and high-level synthesis (HLS) implementation. This work presents a reproducible, bit-exact workflow bridging GNN design and FPGA prototyping for fixed-latency inference. The methodology is demonstrated by implementing a two-layer GraphSAGE network onto an XCVU13P FPGA, using the Cora citation network as a fixed-size
benchmark for firmware evaluation that decouples deployment feasibility from the physics task. Starting from a 32-bit floating-point reference that exceeds the available FPGA resource budget, we derive an integer-only datapath through post-training quantization (INT8 weights and activations, INT32 biases), a power-of-two scale approximation that replaces rescaling multipliers with arithmetic shifts, and data-driven bit-width narrowing. Every stage is validated bit-exactly against a Python integer emulator in Vitis HLS C-simulation. The optimized INT8 power-of-two design achieves an inference latency of 19 clock cycles (52.8 ns at the nominal 360 MHz clock) at 20% DSP, 6% FF, and 27% LUT utilization, with 75.0$\pm$1.1% accuracy compared to the 78.0$\pm$0.8% for FP32. Accuracy is reported as the average across multiple training seeds. The resulting workflow establishes a concrete, transferable path toward fixed-latency GNN-based track reconstruction in the CMS Level-1 trigger.
Read the full abstract ▾
All authors Andrea Cardini, Pelayo Leguina, Elena Aller, Santiago Folgueras
Combinatorics
Sharp Bounds for Kulli-Basava Indices of Graphs
Sanju Vaidya and Jeff Chang
In this paper, we establish formulas and sharp bounds for general Kulli-Basava indices and characterize graphs that attain these bounds. These indices have been shown to possess strong discriminating power for distinguishing nonisomorphic chemical structures. They are based on the edge neighborhood degrees of vertices in a graph. We also establish bounds for several classes of graphs, including triangle- and quadrangle-free graphs and graphs with a prescribed clique number. The formulas and bounds depend on the numbers of vertices and edges, the minimum and maximum edge neighborhood degrees, and the first Zagreb index.
Read the full abstract ▾
Lattice Field Theory
A variational framework for variance reduction in lattice field theory
Pietro Butti, Guilherme Catumba, Alessandro Nada et al. (+1 more)
The signal-to-noise problem limits the reach of many
lattice calculations. We present a variational framework that recasts it as a transport problem: the loss of signal reflects a mismatch between the distribution one samples and the one needed to measure an observable, and can be reduced by transporting configurations to close that gap. The optimal transport is typically determined either through a
stochastic estimator based on Langevin dynamics or by parametrising it as a normalising flow trained with automatic differentiation. We discuss how the framework brings these methods under a common variational principle and present results for scalar theories.
Read the full abstract ▾
All authors Pietro Butti, Guilherme Catumba, Alessandro Nada, Louis Spatscheck
Representation Theory
Locally analytic representations in mixed characteristic via stacks
Maximilian Hauck
For any $p$-adic Lie group $G$, we construct a group stack $G^\mathrm{la}$ over the $p$-adic branch of Clausen–Scholze's stack of norms such that quasicoherent sheaves on its classifying stack $*/G^\mathrm{la}$ recover locally analytic representations of $G$ after base change to any suitable Tate Huber pair, even in positive or mixed characteristic. We also show that smooth $\mathbb{F}_p$-representations of $G$ embed fully faithfully into sheaves on the mod $p$ fibre of $*/G^\mathrm{la}$ and establish Poincaré duality for
cohomology of locally analytic representations in mixed characteristic.
Read the full abstract ▾
History and Overview of Mathematics
Training Mathematicians in the Age of AI: Intellectual Agency, Cognitive Offloading, and the PhD Thesis
Thomas Koberda
Powerful artificial intelligence is weakening the traditional relationship between mathematical output and evidence of mathematical expertise. In particular, the production of an original
theorem or a polished dissertation can no longer, by itself, certify the intellectual formation of its nominal author. I argue that graduate mathematical education should therefore be organized around the formation of intellectual agency: internal technical competence, mathematical judgment, understanding, and responsible participation in a shared intellectual culture. I distinguish productive from premature cognitive offloading, propose complementary independent and AI-augmented modes of training, and suggest a corresponding reformulation of the role of the PhD thesis and dissertation defense. More broadly, I argue that academic mathematics should understand itself increasingly as an institution for the reproduction and stewardship of human mathematical expertise rather than primarily as a mechanism for producing
theorems.
Read the full abstract ▾
Quantum Physics
Beyond Antibunching: Photon Correlation Analysis Reveals Blinking Origin
Patricia Kallert, Santiago Bermúdez-Feijóo, Giorgio De Pascalis et al. (+18 more)
A variety of
quantum emitters compete in the quest for the best single-
photon source for photonic
quantum technologies. Consequently, only a consistent approach to analyze the second-order auto correlation function allows comparison of the multi-
photon contribution ($g^{(2)}(0)$) of different sources. However, the community employs different and inconsistent methods for blinking sources, leading to incomparable
benchmarks of source quality. Here, we use the emission of an inherently non-blinking
quantum dot (QD) and apply artificial blinking through two different mechanisms: masking the recorded raw data in post-processing and emulation of a blinking system by gating the
laser excitation pulses. We then compare the $g^{(2)}(0)$ values with the non-blinking result. For the analysis, we investigate five
estimators of $g^{(2)}(0)$ actively used in the literature. While fitting the envelope of the correlations on long-time scales with the correct blinking model is the best choice, normalizing to the Poisson level gives by far the worst proximity. Furthermore, we test our predictive model to identify the underlying blinking mechanism of a QD in a circular Bragg grating cavity.
Read the full abstract ▾
All authors Patricia Kallert, Santiago Bermúdez-Feijóo, Giorgio De Pascalis, Nicolas Claro-Rodriguez, Ioannis Caltzidis, Normen Auler, Eva Berger, Rebecca Aschwanden, Tobias Krieger, Saimon. F. Covre da Silva, Sandra Stroj, Quirin Buchinger, Michele B. Rota, Thomas Hummel, Sven Höfling, Armando Rastelli, Rinaldo Trotta, Dirk Reuter, Sonja Barkhofen, Klaus D. Jöns, Tobias Huber-Loyola
Atomic Physics
Sub-Doppler cooling of trapped ions using optical tweezers
Bas Gerritsen, Liam J. Bond, Jiri Minář et al. (+3 more)
We propose a sub-Doppler cooling scheme for trapped-ion
crystals in Paul traps. The combination of a hollow tweezer and microwave drive creates an effective two-level system which is cooled through attractive Sisyphus cooling. For ${}^{171}\rm{Yb}^+$ ions, we numerically identify optimal parameters for one- and two-ion
crystals, and investigate the cooling rate in ion
crystals with up to three ions. Our cooling scheme utilizes a single ion addressed with a tweezer to achieve sub-Doppler cooling of the $N$ axial motional modes perpendicular to the tweezer propagation direction, does not rely on the Lamb-Dicke regime and uses off-resonant light. We reach $1/e$ cooling times for the center of mass mode of $\tau_{\rm com}<2$ ms, and a steady-state excitation number of $\bar{n}_{\rm{com}}\approx 0.7$. Our scheme is well-suited as an intermediate stage between Doppler cooling and ground-state sideband cooling, and can be readily realized in trapped-ion platforms.
Read the full abstract ▾
All authors Bas Gerritsen, Liam J. Bond, Jiri Minář, Arghavan Safavi-Naini, Robert J. Spreeuw, Rene Gerritsma
Computers and Society
"We'll Fix It Later": Education, AI, and the Deferral of Privacy in EdTech
Meghna Manoj Nair and Rachel Greenstadt
Educational technology (EdTech) platforms collect highly sensitive student data, including behavioral logs, disability records, and academic histories. However,
privacy considerations are often postponed rather than treated as a foundational design requirement. We present a mixed-methods study combining 12 semi-structured interviews with EdTech professionals and a
privacy policy audit of 48 platforms coded across five dimensions, with strong inter-rater reliability (mean Cohen's Kappa = 0.781). Our interviews reveal a recurring organizational pattern in which
privacy is recognized as important but deferred across the product lifecycle as organizations prioritize product functionality, growth, funding, and immediate educational outcomes. Responsibility is often delegated to cloud providers, policy documents, or downstream institutions, while limited
privacy-related feedback gives organizations little pressure to change these practices. The policy analysis reflects these patterns: platforms describe what data they collect relatively well but provide substantially less information about how that data is subsequently governed. Thirty-three percent make no meaningful Artificial Intelligence (AI) disclosure despite visible AI features, and 73% provide only generic accountability and breach-response
language. K-12 platforms perform better on children's consent where regulation creates explicit requirements, but this advantage does not extend to AI governance or accountability. These findings suggest that meaningful improvement requires enforceable institutional and regulatory mechanisms rather than voluntary
privacy commitments alone.
Read the full abstract ▾
Quantum Physics
Unbounded Operators, Physicality, and Modality in Quantum Theories
Zhonghao Lu
In this paper, I address the problem regarding the physicality of
Hilbert space in
quantum mechanics, a topic brought into focus by Carcassi et al.'s paper "The unphysicality of
Hilbert spaces" recently. I argue for the physicality of the
Hilbert space in
quantum mechanics, and demonstrate that there are inevitable ambiguity and arbitrariness if we want to identify the set of physical states with a proper subset of it. Moreover, I distinguish the ontological and modal reading of physicality from the pragmatic and operational reading of physicality, and then develop what I call the hierarchical view of physicality. According to this view, physicality is not a monolithic concept, but a layered one that admits degrees. I show the connection between the concept of physicality and modal metaphysics, and finally explain how it can be used to address the interpretational problem of
quantum field theory brought by the existence of inequivalent representations.
Read the full abstract ▾
Numerical Analysis
A Lloyd-stabilized Voronoi particle method
Bruno Després and Borjan Geshkovski
We study the stability and consistency of Lloyd's
algorithm used in combination with the Lagrangian transport of a density on a Voronoï tessellation. We show that relaxation rates as strong as $O(h^{-1/2})$, where $h$ is the mesh size, give convergence to the continuity equation in Wasserstein distance at the rate $O(h^{1/4})$. Thus the mesh can be kept regular by the correction alone, without the remeshing that Lagrangian methods usually require. An application is proposed for the compressible Euler equations.
Read the full abstract ▾
High Energy Astrophysical Phenomena
Variable opacity and accretion-column regimes of magnetized neutron stars
Galina Lipunova and Pavel Abolmasov
Radiation-supported
accretion columns of magnetized
neutron stars can operate in several distinct regimes. As the
accretion rate increases, magnetospheric
accretion may proceed through hot-spot emission, efficiently cooling radiative shocks, and advective sinking columns. The structure of the
accretion flow is determined by the global parameters such as
magnetic field strength and mass
accretion rate, but is also affected by opacity variations in strong
magnetic fields. A decrease of opacity for
photons with energy below the cyclotron energy is able to substantially augment the classical critical
accretion rate separating efficiently cooling solutions from advective solutions. We propose an approach to determine the critical
accretion rate above which the Basko-Sunyaev sinking solution becomes unavoidable. We also use the Basko-Sunyaev sinking solution to estimate the
plasma temperature and compare the characteristic
photon energy with the local cyclotron energy. For a combination of moderate
magnetic fields and high
accretion rates, the characteristic
photon energy in advective columns typically exceeds the cyclotron energy. In this case, the opacity is not reduced by the
magnetic field, and tall
accretion columns are formed. Strong
magnetic-field systems may retain short, efficiently cooling
accretion columns over a wide range of super-Eddington
accretion rates. We identify a region in
accretion rate and
magnetic-field strength where both efficiently cooling and advective solutions may coexist. In this region, the column may switch or oscillate between the regimes and geometries. The typical time of such oscillations are close to the replenishment time of the column, or to the thermal time near its bottom. The quasi-periodic oscillations, observed in some super-Eddington objects in the 1-100 mHz frequency range, may be related to such relaxation cycles.
Read the full abstract ▾
Quantum Physics
Optimal local oscillators for the homodyne detection of multiphoton states
Gisell Lorena Osorio, Paul Virally, Sean Molesky et al. (+1 more)
We propose a framework for optimizing pulsed local oscillators (LO) for homodyne detection of multiphoton-number states by exploiting the tensor structure of their joint-spectral amplitude (JSA). We show that finding the optimal LO is equivalent to computing the JSA tensor's leading
unitary eigenpair and that the factor
matrices of the JSA's Tucker higher-order singular value decomposition (HOSVD) coincide with the Schmidt modes of the
photon number state's single-particle reduced density
matrix. We use the HOSVD to bound the optimal homodyne visibility and to initialize gradient based optimization. In simulated JSAs, weakly correlated, few-mode states reach near-unity visibility, with the leading HOSVD mode being the optimal LO, while strongly correlated, multimode states require full optimization and saturate below unit visibility even at the true optimum. These results offer a practical route to design LOs for homodyne detection experiments with realistic sources of
photon-number states, which contributes to the practical implementation of sources of non-Gaussian
quantum states.
Read the full abstract ▾
All authors Gisell Lorena Osorio, Paul Virally, Sean Molesky, Nicolás Quesada
Physics and Society
Bifurcation and Multistability in a Collective-Risk Public Goods Game with Quorum-Activated Protection
Shulan Li, Xiaogang Li, Jun he et al. (+2 more)
The danger of collective risk in a public goods game could provide an effective escape route from the tragedy of the commons. Existing studies usually embed risk mitigation in the contribution itself, whereas many real systems rely on a separate costly and non-excludable protective activity that requires a participation quorum. Whether such protection can evolve and persist remains unclear. Here, we formulate an $N$-player public-goods game with ordinary cooperators, defectors, and protective cooperators. Defectors endogenously increase a saturating probability of collective failure, whereas protective cooperators incur an additional cost and reduce this risk for all group members once their number reaches the quorum. Replicator analysis and numerical bifurcation calculations yield three main findings. First, protective cooperation gains an advantage when the focal participant completes the quorum and the resulting expected loss reduction exceeds its additional cost. This advantage varies non-monotonically with group composition. Second, when a single protective cooperator cannot activate protection, protection effectiveness does not enter its first-order invasion fitness at full defection; greater effectiveness cannot overcome the rare-invasion barrier when full defection is locally asymptotically stable. Third, a boundary saddle–node bifurcation creates a saddle and a locally asymptotically stable coexistence
equilibrium of defectors and protective cooperators. When full cooperation and full defection are also locally asymptotically stable, three attractors coexist, and the numerically observed evolutionary outcome depends on the initial population composition. These results distinguish the effectiveness of activated protection from its ability to become established from rarity.
Read the full abstract ▾
All authors Shulan Li, Xiaogang Li, Jun he, Lei Shi, Attila Szolnoki
Robotics
DEAL-Grasp: Decoupled Alignment Representation for Geometry-Aware Dexterous Grasp Generation
Fuqiang Zhao and Qian Liu
Synthesizing realistic articulated hand-object interactions is a fundamental problem in virtual reality,
embodied intelligence, and digital human applications. Existing methods for dexterous grasp synthesis typically regress or denoise poses in a joint space that couples global rigid motion with local articulation, which often yields unstable samples and physically implausible contacts. We introduce DEAL-Grasp, built upon the Decoupled
Alignment (DEAL) representation, which reformulates grasp synthesis as
alignment-space generation: the interaction state comprises task-space
geometric anchors and articulation parameters, from which the rigid transform is recovered via closed-form Procrustes
alignment while preserving local articulation. On this mixed state, we model grasp generation using heterogeneous-state flow matching with component-wise vector fields, incorporating time-adaptive physical regularization during training. At inference, grasps are synthesized solely by integrating the learned vector field, without test-time optimization or auxiliary physical guidance. Across MultiDex and zero-shot RealDex
benchmarks, DEAL-Grasp attains high force-perturbation success rates alongside minimal penetration and high diversity of generated grasps, while substantially reducing native inference latency compared to optimization-heavy baselines. The project page is available at https://wmtlab.github.io/DEAL-Grasp/.
Read the full abstract ▾
Software Engineering
Scenario-Driven Neuroevolution: Using Models to Guide Test Generation for Games
Gijs van Cuyck, Patric Feldmeier, Jan Tretmans et al. (+1 more)
Automatically generating test inputs for games is challenging, as test generators must master the game to reach advanced program states while also ensuring robustness against the heavy program randomisation inherent to games. The test generator Neatest therefore optimises test suites consisting of neural networks that reach advanced program states and are robust to program randomisation, as they generate test inputs dynamically based on the current program state. Neatest is a white-box testing approach that aims to generate a network
agent for each yet-uncovered statement or branch of the code using neuroevolution. Due to this iterative test generation approach, the
algorithm does not scale well to larger programs that may contain thousands of branches. Furthermore, covering every statement or branch in a game often does not correspond to playing the game as intended. To alleviate these shortcomings, we propose combining Neatest with a model-based testing approach that allows game testers to define test scenarios via abstract game models. The test generator then no longer optimises networks to reach all branches or statements of a program, but instead trains networks to replicate the concrete desired testing behaviour expressed by the abstract game model. An evaluation on 13 Scratch games across varying genres demonstrates that Neatest, combined with model-based testing, is able to optimise
agents that replicate the desired gameplay behaviour defined in the game models while increasing achieved branch coverage by 7% compared to the traditional code-guided Neatest approach.
Read the full abstract ▾
All authors Gijs van Cuyck, Patric Feldmeier, Jan Tretmans, Gordon Fraser
High Energy Physics (Phenomenology)
Environment-dependent mass splitting to suppress solar capture in the inelastic-doublet interpretation of the LZ event
Sudhakantha Girmohanta and Tae Hyun Jung
The inelastic electroweak-doublet interpretation of the $248\,{\rm keV}$
recoil energy event reported by LUX-ZEPLIN faces strong constraints from
solar capture. We propose an environment-dependent mass splitting generated by an ultralight scalar coupled quadratically to electrons. Its density-induced expectation value enhances the splitting in the Sun while preserving the laboratory value, providing a mechanism to suppress
solar capture. We derive conditions for this mechanism and examine constraints from
stellar cooling and electron-mass variations. We find that this scenario is phenomenologically viable, while it suffers from severe fine-tuning of the scalar potential.
Read the full abstract ▾
Optimization and Control
Inverse knapsack at two capacities: which pairs of value-cardinality hulls are realizable?
Prashant Chaudhary and Kapil Khandelwal
One item set evaluated at two capacities $R<D$ produces two concave hulls of optimal value against cardinality. We ask which prescribed pairs arise. Exchange arguments give a necessary system on vertex witnesses, exchange closure (EC), whose scalar consequences form the linear closure. We exhibit a pair satisfying every scalar test that fails EC, so the linear closure is strictly larger already at larger terminal count three; and a globally coherent EC witness admitting no common-size representation although its target pair is realizable. Under a cardinality cap, a four-band classification of one family gives exact thresholds for cap-four realizability, uncapped realizability and the vertex-only capped closure. Pairs whose larger terminal count is at most two are characterized. Under the explicit encoding the decision problem lies in $\Sigma_2^p$ and is
polynomial-time for fixed terminal cardinalities. Exact finite certificates establish agreement of the scalar and witness conditions on six specified domains; sufficiency of uncapped EC remains open.
Read the full abstract ▾
Statistical Methodology
Extreme Population Selection under Multistage Sampling design With Applications
Shivam, Bhargab Chattopadhyay and Nil Kamal Hazra
We study the problem of selecting the extreme (best or worst) population from among $K(\geq 2)$ populations, under the assumption that the extreme population is sufficiently separated from the nearest population. The selection is based on an appropriate measure, which may vary across different application domains. Since the actual value of the measure is unknown, we obtain its
estimator using the generalized method of moments under a multistage sampling design. Using this
estimator, we propose two sequential procedures, namely online
algorithm and multi armed bandit based
algorithm. Under suitable regularity conditions and without imposing parametric assumptions on the underlying distributions, both
algorithms correctly identify the extreme population with a desired level of confidence. We illustrate the proposed
algorithms through applications in econometrics and genetics. In the econometric application, the extreme population is selected using the Gini index as a measure of inequality and the performance of the proposed procedures is assessed through extensive Monte Carlo simulation studies conducted under various distributional settings. In the genetics application, the worst population is identified using a measure derived from the tumor mutation burden (TMB) score and the practical applicability of the proposed
algorithms is demonstrated using the Memorial Sloan Kettering-IMPACT 50000 clinical sequencing cohort. Further, we use the proposed framework to identify an anomalous population, provided such a population exists.
Read the full abstract ▾
Statistical Mechanics
Optimal Evasion from a Resetting Searcher
Ami Taitelbaum, Matan Shporer, Shlomi Reuveni et al. (+1 more)
Returning to a familiar location can help a searcher find its target—but how should the target move to avoid being found? We address this question for a resetting Brownian searcher and a mobile target fluctuating around a home location. Keeping the target's typical exploration range fixed, we find that the mean capture time is maximized at an intermediate correlation time: rapid motion promotes encounters, whereas slow motion leaves the target effectively stationary. Between these extremes, persistent excursions away from the searcher's reset point can substantially delay capture. Analytical limiting results explain this optimum, while numerical solutions reveal an approximately linear relation between the maximal mean capture time and the optimal correlation time across many decades. These findings identify temporal correlations as a resource for evasion, allowing a target to prolong survival by changing how it moves rather than where it roams.
Read the full abstract ▾
All authors Ami Taitelbaum, Matan Shporer, Shlomi Reuveni, Michael Assaf
Mesoscale and Nanoscale Physics
Helical Anomaly from Concurrent $\mathbb{Z}$ and $\mathbb{Z}_2$ Topology in an Acoustic Semimetal
Hou-Yin Li, Hua-Shan Lai, Jian-Lan Xie et al. (+3 more)
Topological classification constitutes a cornerstone of modern condensed-matter physics. Historically, topological systems have been predominantly characterized by invariants confined to a single topological class, such as topological
insulators with a $\mathbb{Z}_2$ invariant and Weyl semimetals with an integer ($\mathbb{Z}$) topological charge. While such classification frameworks in principle permit gapless phases in which bulk and boundary degeneracies are protected by distinct invariants, this regime has received limited focused
attention, as coordinating the requisite
symmetries within a single band structure is highly nontrivial. Here, we realize such concurrent $\mathbb{Z}$ and $\mathbb{Z}_2$
topology in a two-dimensional acoustic semimetal, revealing pseudospin-selective topological bulk modes arising from the interplay between these distinct invariants, termed helical anomaly bulk states (HABSs). Our
lattice design exploits layer, sublattice, and gauge-staggered degrees of freedom, enabling modular control of effective time-reversal,
chiral, and particle-hole
symmetries. The resulting system exhibits bulk degeneracies protected by an integer $\mathbb{Z}$ invariant and Kramers-like boundary degeneracies protected by a $\mathbb{Z}_2$ invariant, forming twisted boundary arcs. Crucially, finite-size systems manifest helical or anti-helical edge states coexisting with HABSs, a distinctive signature absent in single-invariant
topologies. Our results establish HABSs as a characteristic signature of concurrent
topology and demonstrate a new regime where bulk and boundary invariants jointly govern topological transport beyond conventional single-class frameworks.
Read the full abstract ▾
All authors Hou-Yin Li, Hua-Shan Lai, Jian-Lan Xie, Xiao-Chen Sun, Yan-Feng Chen, Cheng He
Soft Condensed Matter
Active Self-Consistent Field Theory for Ornstein-Uhlenbeck Polymers
Yuliang Huang, Chun-Lai Ren and Qiyun Tang
We develop an active self-consistent field theory (ASCFT) for studying the steady-state behavior of active Ornstein–Uhlenbeck
polymers. Starting from the
stochastic equations of motion under the unified colored noise approximation, we derive an effective
Hamiltonian that extends the classical
polymer field theory to non-
equilibrium systems. The resulting free energy functional incorporates both the Flory–Huggins interaction parameter $\chi N$ and the persistence time $\tau$ of the active noise, enabling a unified description of thermodynamic and activity-driven effects. To solve the governing equations, we implement a stable implicit-explicit numerical scheme that handles the fourth-order term induced by activity. Our simulations reveal that increasing activity suppresses microphase separation, with the density modulation amplitude decaying as $\Delta\phi \propto \tau^{-1/2}$ for large $\tau$. This scaling is independent of copolymer composition and is confirmed by an asymptotic analysis of the free energy functional in the large-$\tau$ limit. The ASCFT framework provides a new theoretical tool for predicting and designing the non-
equilibrium morphologies of active
polymer systems, bridging the gap between traditional self-consistent field theory and active matter physics.
Read the full abstract ▾
General Topology
Scott–Isbell Coincidence for Continuous Dcpos beyond Bicompleteness
Yuxu Chen, Yunjie Wen and Xiaoyong Xi
Lawson and Mislove posed the following problem in 1990 as Problem~535 in Open Problems in
Topology: for a core-compact space \(X\) and a dcpo \(P\) equipped with its Scott
topology, under what conditions on \(P\) do the Isbell and Scott
topologies on \(C(X,P)\) agree?It was proved that, for a nonempty bicomplete continuous dcpo \(P\), the Isbell and Scott
topologies on \(C(X,P)\) coincide for every core-compact space \(X\) if and only if \(P\) is bounded complete; for every compact core-compact space \(X\) if and only if \(P\) is conditionally bounded complete; and for every RW-space \(X\) if and only if \(P\) is a pointed continuous \(L\)-domain.We remove the bicompleteness assumption from all three classifications by combining the forbidden-retract
theorem of Jia, Jung and Li with a separation
theorem for powers of downward well-ordered chains and suitable Alexandrov test spaces.
Read the full abstract ▾
Machine Learning (Statistics)
NPBoost: Neural Processes with Gradient-Boosted Fixed Effects
Andrea Nava, Ken Rölli, Armin Begic et al. (+1 more)
Neural Processes (NPs) are model-based meta-learners that implicitly learn a
stochastic process and adapt to a new task from a small context set. Most extensions of NPs focus on improving the neural network architecture. We instead develop an extension motivated by the shared hierarchical interpretation of meta-learning and mixed-effects models. Specifically, we introduce Neural Process Boosting (NPBoost), which decomposes structured response variability into tree-boosted fixed effects shared across tasks and NP random effects that capture
stochastic task-to-task variation. We propose to train the two components jointly using a boosting
algorithm in which an NP learns residual task-specific structure and a tree ensemble estimates common patterns across tasks. Across synthetic and real-world tabular meta-learning problems, this decomposition improves over a standard NP when the shared structure contains discontinuities or other irregular patterns that boosted trees can represent effectively.
Read the full abstract ▾
All authors Andrea Nava, Ken Rölli, Armin Begic, Fabio Sigrist
Numerical Analysis
An efficient 0D-space conservative and positivity preserving battery model
S. Clain, M.C. Lucas, M.T. Malheiro et al. (+1 more)
Embedded software aims to provide the main characteristics of a
cell in real time to manage and optimize battery usage. A complete multidimensional model is not tractable with the low computational resources available in-situ. A class of 0D-in-space models is derived from the P2D Doyle-Fuller-Newman system, with additional assumptions about the functions' shape in space, and strongly reduces the complexity while preserving the main characteristics over time. We propose a conservative, positivity-preserving model built from the P2D system based on the statement that the Butler-Volmer source term is constant in each subdomain (anode, separator, cathode). We develop a mathematically rigorous construction that provides the simplified model and design a scheme to solve the system. We highlight that the model guaranties the conservation property together with the positivity of the concentrations. In particular, the
diffusion-limited current density that represents the maximum current that can be used while still preserving the positivity of the cation concentration is obtained explicitly.
Read the full abstract ▾
All authors S. Clain, M.C. Lucas, M.T. Malheiro, C. M. Costa