Game Theory
Papers
Fresh from arXiv, typeset properly. Highlighted words trade on the market.
All of arXiv
Monday 31 August
We study fair allocation of indivisible goods under additive valuations and matroid constraints. A challenging open question is whether a complete and feasible envy-free up to one good (EF1) allocation exists under every matroid that admits a complete and feasible allocation. The state-of-the-art result by Biswas and Barman [2018] positively resolves this question for partition matroids. Our first result positively resolves it for laminar matroids, which generalize partition matroids, when there are three agents. Our technique reduces this general existence question to finding an EF1 allocation satisfying a mild additional condition under a single finite-sized key laminar matroid, and we establish the required allocation by case analysis. We show that our technique somewhat extends to four agents, reducing the analogous problem to finding EF1 allocations under two finite-sized laminar matroids, although we are unable to establish their existence. We also use recent matroid decomposition results to establish EF1 existence under broader classes of matroids. Specifically, we show that EF1 allocations always exist under transversal matroids whenever a complete allocation is feasible, and obtain existence results for graphic matroids and gammoids under stronger assumptions.
Machine Learning
On the Complexity of the Compatibility Problem for Succinctly Encoded Conditional Distributions
The motivation for this paper is the investigation of the trade-offs implicit in probabilistic models used in machine learning. Models are often used to make predictions in the form of conditional probabilities. However, a pair of conditional distributions p(x|y) and p(y|x) may not be compatible with any joint distribution p(x,y). Given two such conditionals, determining if there exists a compatible joint is known as the compatibility problem. For discrete random variables, when the conditionals are encoded as probability tables, the compatibility problem has a known solution, which is computationally tractable. In this paper, we formalise and study a succinct version of the problem, encoding conditional distributions as arithmetic circuits. This is applicable to practical applications of probabilistic modelling in high-dimensional settings, including neural network models. We show that, for succinct circuit representations of conditionals, the compatibility problem is intractable. In the case that all probabilities are non-zero, the problem is co-NP-complete. In the case that probabilities can be zero, we give examples to demonstrate that several notions of compatibility can be distinguished, and we prove that multiple versions of the problem are PSPACE-complete. Furthermore, we show that, assuming the polynomial hierarchy does not collapse, there exist compatible succinct conditionals whose joint cannot be expressed succinctly. Implications of these results for probabilistic modelling and machine learning are discussed.
Read the full abstract
Computation and Language
PaperGym: Rubric-Centered Evolution for Research-Plan Generation
Research planning is the decisive capability of AI scientists. Yet a research plan admits no verifiable answer, so reinforcement learning lacks the environment it requires: tasks paired with a critic. Rubrics extracted from scientific papers can supply the critic. Existing pipelines, however, draw the question and the criteria from the same content, so the reward can be earned by paraphrase. The rubric is further compressed into a single scalar per rollout. We introduce PaperGym, a unified framework that turns each research paper into a complete training environment. PaperGym exploits the structure of a paper: the question is synthesized from the research goal and background, while the criteria are derived from the method and experiments. The criteria span methodological innovation and experimental design, and criterion leakage falls to 3.7%, versus 11.90% to 34.10% in existing datasets. Training uses the rubric twice: first as privileged context for OPSD's self-teacher, then as the reward for GRPO. Across Qwen3-1.7B/4B/8B, this schedule outperforms supervised fine-tuning, either stage alone, and the reverse ordering, improving five-benchmark averages by +5.6, +5.0, and +4.8 points. With the recipe held fixed, models trained on PaperGym-20k win 58.1% of three-way comparisons, against 28.2% for RubricHub Science. The trained Qwen3-8B reaches 73.48 on ResearchQA, above the far larger Kimi K2.6. We release the pipeline, the 20,000-instance corpus PaperGym-20k, and the benchmarks PaperGym-Innov and PaperGym-Design.
Read the full abstract
Artificial Intelligence
When Does Bigger Help? A Controlled Study of LLM Scale for Ontology Learning
The effect of Large Language Model (LLM) scale on ontology learning (OL) performance remains insufficiently characterized. We present a controlled evaluation of 13 models spanning dense and Mixture-of-Experts variants from the Qwen3.5 and Qwen3.6 lineages, together with proprietary GPT release variants, using the OntoLearner retrieval-augmented generation pipeline. All models are evaluated with the same embedding model, retrieval configuration, prompt templates, decoding settings, datasets, and metrics on term typing, taxonomy discovery, and non-taxonomic relationship extraction across four biomedical and materials science and engineering ontologies. Within the dense Qwen3.5 lineage, increasing parameter count primarily improves precision rather than recall, with the largest gains occurring between 9B and 27B parameters. However, the effect of scale is neither monotonic nor uniform across tasks and domains. Dense 27B models outperform substantially larger sparse models on term typing, whereas larger Mixture-of-Experts models achieve the strongest open-weight results on taxonomy discovery. Non-taxonomic relationship extraction remains difficult across model scales, particularly for the Materials Data Science ontology. Performance differences across matched Qwen variants and proprietary GPT releases further indicate that architecture and model lineage can outweigh nominal parameter count. These findings show that model size alone is an insufficient selection criterion for OL and provide empirical guidance for reproducible LLM-assisted ontology engineering.
Read the full abstract
Quantum Physics
"Train classical, deploy quantum" requires rethinking generalization
Generative models have become central across science and industry, from image and text synthesis to the design of molecules and materials. Quantum generative models are considered one of the most promising applications for quantum computers, since a quantum circuit naturally produces samples from the distribution it encodes, and for suitable circuits that distribution is believed to be hard for any classical computer to reproduce. A leading strategy trains these models on a classical computer and reserves the quantum device for generating samples at deployment. This is possible when the training loss can be evaluated on a classical computer. A prime example is the maximum mean discrepancy (MMD$^2$), a moment-matching loss that compares the model and the data through their Pauli-$Z$ correlations. Research so far has asked whether such models can be trained and whether their sampling is hard; whether minimizing such an objective yields a model that generalizes, rather than one that merely reproduces the training statistics, remains poorly understood. We benchmark thirteen quantum and classical generative models by direct sampling on two application-inspired datasets: first a cardinality-constrained dataset at up to $30$ qubits and second a dataset of genomic single-nucleotide variants, whose valid set is the observed data. Models that converge the loss to the same value differ widely in how much of the unseen valid set they cover. These results indicate that a converged moment-matching loss is not a reliable measure of generalization, and that a train-classical, deploy-quantum workflow has to measure generalization by sampling the trained model, a step that at the sizes of interest is believed to require the quantum device.
Read the full abstract
Quantum Gases
Static-Field Shielding of Bosonic Molecules: Evaporation to Degeneracy and Self-Bound Droplets
The strong, tunable dipolar interactions of ultracold molecules make them a powerful platform for quantum many-body physics, but reaching degeneracy by evaporative cooling requires suppressing inelastic collisions. Microwave collisional shielding has enabled the preparation of degenerate Fermi gases and Bose-Einstein condensates of polar molecules, whereas static electric field shielding has been limited to fermionic species, which are less prone to inelastic loss. We demonstrate static-field Förster shielding of bosonic $^{23}$Na$^{87}$Rb molecules, suppressing two-body loss by up to four orders of magnitude. Within a bound-state-free electric-field window, three-body loss is also strongly suppressed, enabling efficient evaporation. Evaporating with an efficiency of 2.09(9), we increase the phase-space density of the gas by two orders of magnitude, reaching degeneracy with 7200(1000) molecules. We observe self-bound droplets at the end of evaporation over a wide range of field strengths, emerging from either degenerate or non-degenerate parent gases. We perform Path Integral Monte Carlo simulations, which suggest that the observed droplets are filamentary in nature, and find good agreement with the experimentally observed droplet formation temperatures. Our results establish Förster shielding as a single-field route to prepare degenerate gases and self-bound droplets of bosonic polar molecules.
Read the full abstract
Human-Computer Interaction
InsightToast: Proactive Information Retrieval & Glanceable Visualization in the Side Channel of Data-Rich Meetings
Missing institutional context during meetings can impede effective participation. Retrieving relevant information, often scattered across heterogeneous internal and external sources, requires costly task-switching that disrupts both individual focus and collective conversational flow, particularly detrimental during cognitively demanding tasks such as decision-making. We introduce InsightToast, a mixed-initiative application that monitors verbal discourse in real time, identifies topics and informational needs as they emerge, and proactively retrieves relevant information through a multi-agent large language model (LLM)-based pipeline integrating retrieval-augmented generation (RAG) to produce source-grounded insights as succinct text and glanceable interactive charts, delivered through a peripheral interface as ephemeral toasts in the conversation's side channel. To demonstrate the potential for yielding serendipitous insights, we showcase a usage scenario involving a knowledge base of legislative documents as the meeting's context. We then report on a comparative study (N=16), in which participants arrived at informed policy decisions while maintaining natural conversation flow.
Read the full abstract
Statistical Mechanics
Overcoming critical slowing down in frustrated spin systems by learned multiscale sampling
Cluster algorithms, such as the Swendsen–Wang and Wolff methods, are among the most successful MCMC methods for mitigating critical slowing down in statistical systems. These constructive cluster algorithms, however, fail in the presence of even extremely weak frustration. Here, we sidestep this fundamental limitation by learning rather than constructing the relevant clusters. Specifically, we use the wavelet conditional renormalization group (WCRG) sampling method to learn the probability distribution of collective fluctuations of a frustrated two-dimensional soft-spin model. Configurations are then generated recursively from coarse to fine scales by sampling conditional wavelet distributions. The WCRG method reproduces the main statistical properties of the system across different phases, including the local-field distribution and the structure factor. At an Ising-like critical point, the conditional dynamics remains decorrelated within $\mathcal{O}(1)$ sweeps at each scale, yielding an overall sampling complexity of $\mathcal{O}(\log_2 L)$, thus making WCRG much more efficient than standard local MCMC methods. These results show that learned multiscale sampling can overcome critical slowing down in frustrated systems for which conventional cluster algorithms fail. By assessing the sampling accuracy of different observables, we also clarify the main tradeoff of the WCRG method: the accuracy of the fast sampling scheme depends on the expressiveness of the energy-based model used to estimate the wavelet conditional distributions.
Read the full abstract
Computer Vision
BLARM: Animating 3D Objects from Video via Blending Latent Rigid Motion Primitives
We introduce BLARM, a feed-forward method for video-driven 3D mesh animation. Given a monocular video and a static object mesh, BLARM predicts a temporally coherent animated mesh whose motion follows the video. Rather than relying on explicit rigs or directly regressing high-dimensional vertex motion, we represent animation using a compact set of learned, time-varying rigid motion components and time-invariant vertex-to-component skinning weights. This yields a low-dimensional deformation space without requiring skeletons, cages, skinning weights, or rig annotations. Our architecture conditions geometry-derived deformation latents on video features through factorized spatial-temporal attention, then decodes rigid transformations blended by predicted skinning weights. Trained with trajectory reconstruction, entropy regularization, and motion-aware contrastive learning, BLARM produces accurate and temporally stable animations while recovering compact, interpretable motion structure from monocular video.
Read the full abstract
Quantum Physics
Unconditional Certified Randomness without Structure
We obtain a certified randomness protocol in the quantum random oracle model. The protocol is non-interactive and publicly verifiable with a classical verifier, and is based on Yamakawa and Zhandry's proof of quantumness [JACM'24]. We prove unconditional security of this protocol against adversaries making subexponentially-many adaptive quantum queries to the random oracle. Prior work on certified randomness relative to a random oracle additionally assumed the Aaronson–Ambainis conjecture or proved security only against low query-depth adversaries.
Read the full abstract
Computation and Language
Aspire: Can Models Self-Evolve from Vague Goals?
Many important forms of human learning begin with a vague goal, such as "become a better physicist" or "improve at research." Learners must interpret the goal, identify capability gaps, decide how to learn, and determine whether they have actually improved. In contrast, existing work on LLM self-evolution typically begins with tasks and evaluation metrics specified by humans, reducing self-evolution to optimizing an explicit objective rather than deciding what and how to learn. We introduce ASPIRE, a benchmark for vague-goal-driven self-evolution. ASPIRE provides only a natural-language capability goal while downstream evaluation tasks remain hidden. The agent must operationalize the goal by choosing data and update methods, constructing training and validation signals, and deciding when to evaluate. ASPIRE supports both model-weight and agent-harness evolution in a unified interactive environment and evaluates the resulting systems on a hidden, expert-authored set of 520 items spanning six goals. Our experiments show that vague goals redirect search effort toward goal interpretation. Current agents routinely complete training and harness-editing loops, but weight-level gains remain sparse and unstable, and the strongest evolved harness remains below the engineered Qwen-Agent reference. Agents often train on mismatched data and trust narrow self-evaluations, so local gains fail to transfer to hidden evaluation and continued search and training can erase earlier improvements.
Read the full abstract
Solar and Stellar Astrophysics
Accretion Rates and Thermohaline Convection in Polluted White Dwarfs
The atmospheres of polluted white dwarfs have information about the composition of disrupted planetary material. However, interpreting the observed abundances requires a quantitative description of the mixing processes that transport such material through their envelopes. The accretion of metal-rich debris drives thermohaline convection in the star's radiative zone. Previous studies have modeled this form of turbulence as an enhanced diffusive process, often inferring mass accretion rates $\dot{M}_{\rm acc}$ that exceed X-ray constraints. We develop a time-dependent model that treats thermohaline mixing as a propagating front whose evolution is coupled self-consistently to the surface abundance of heavy elements. We perform hydrodynamical simulations which demonstrate that thermohaline mixing drives the composition gradient towards values slightly above marginal thermohaline stability. Using this condition, we derive analytic solutions for the mixed layer depth $h$ and surface abundance $X_{\rm surf}$, finding that both scale as $t^{1/2}$ at early times, transitioning to $h \propto t^{4/27}$ and $X_{\rm surf} \propto t^{10/27}$ at later times. Without gravitational settling, the thermohaline front never reaches a steady state, while including settling leads to an equilibrium surface abundance $X_{\rm surf, eq} \propto \dot{M}_{\rm acc}^{30/43}$. Applied to G 29–38, our model matches the observed heavy-element abundance for $\dot{M}_{\rm acc} \sim 1\times 10^{10}$–$2\times 10^{10}~\mathrm{g~s^{-1}}$, comparable to previous findings using stellar evolution codes. These rates exceed the X-ray estimate by about an order of magnitude, which may point to a partially suppressed X-ray luminosity from the accretion boundary layer. We present a practical prescription for inferring accretion rates directly from the observed surface abundance and stellar parameters of other polluted white dwarfs.
Read the full abstract
High Energy Physics (Phenomenology)
Columnar Recasting and High-Luminosity LHC Projections for a Light $L_\mu - L_\tau$ Gauge Boson in the Four-Muon Channel
We present an independent recasting of light vector gauge boson ($Z'$) searches at the LHC within the leptophilic $U(1)_{L_\mu - L_\tau}$ framework. A simplified Standard Model extension is implemented in FeynRules, enforcing symmetric chiral leptonic currents while suppressing hadronic couplings. Signal generation for the four-muon ($4\mu$) final state is performed using MadGraph5_aMC@NLO, Pythia8, and a tuned Delphes3 detector emulation, covering the mass range 5-62 GeV under a reference coupling g=0.01. The analysis employs a vectorized columnar pipeline in Coffea with Awkward Array masks, enabling parallelized kinematic selections and the recovery of boosted muon topologies. Signal yields are statistically extracted via pyhf against digitized CMS Run 2 backgrounds (77.3 fb$^{-1}$). The resulting expected median limit profile tracks the absolute experimental sensitivity morphology within a factor of $\approx 2.5$, a discrepancy successfully quantified as a systematic consequence of fast-simulation detector limitations and the mass-window binned counting strategy relative to experimental unbinned multivariate profiling, thereby establishing a rigorous and conservative baseline sensitivity floor. Furthermore, we project the discovery potential for the future High-Luminosity LHC (HL-LHC) phase at $\mathcal{L}_{\text{int}} = 3000$ fb$^{-1}$ under the official CERN Scenario II recommendations. We demonstrate quantitatively that the HL-LHC will achieve exceptional expected local statistical discovery significances between 10.22$\sigma$ and 16.61$\sigma$ within the 5-28 GeV mass window, comfortably clearing the discovery threshold and establishing the multi-muon channel as a primary probe for lightweight vector portals in upcoming collider phases.
Read the full abstract
Machine Learning
Stress-Testing Efficient Responsible-AI Evaluation: When Compute Savings Change Benchmark Conclusions
Efficient evaluation changes the protocol used to support claims about model behavior, yet it is rarely tested whether those claims remain stable after the evaluation itself is made cheaper. We stress-test conclusion robustness in responsible-AI benchmarking by evaluating three dense and mixture-of-experts models on BBQ and BBQ-V under seven conditions spanning batching, quantization, benchmark reduction, and their combinations. Rather than treating preserved aggregate accuracy as sufficient, we compare accuracy, bias severity and prevalence, reasoning quality, subgroup behavior, subset-membership stability, runtime, and measured GPU energy against a full-benchmark BF16 baseline. Larger batching keeps accuracy within 0.35 percentage points of baseline and produces comparatively small subgroup changes, while reducing energy in five of six model–dataset settings. INT8 largely preserves quality but uses 1.79–4.26$\times$ baseline energy. INT4 causes larger, model- and context-dependent changes. Reduced benchmarks provide the most consistent savings, but very small subsets are substantially more sensitive to which items are retained. Efficient evaluation should therefore be treated as a measurement intervention whose validity must be checked across the conclusions the benchmark is intended to support. Our project website is https://vectorinstitute.github.io/sustainable-rai-evaluation/ and the code is available at https://github.com/VectorInstitute/sustainable-rai-evaluation.
Read the full abstract
Computer Vision
VeriCam: A Verification Baseline for the Classification of Unknown Data
The advent of foundation models have enabled a new era in zero-shot classification. Yet, key challenges persist. Despite their impressive generalization power that leverages the immense pre-training knowledge, both foundation models for image and text as well as vision-text hybrids lack the representational power needed for fine-grained, minutiae-based class separation that some real-world tasks require. To address the current gaps in the literature, we propose VeriCam, a pipeline designed to learn highly specialized features that enable classification of unknown classes in unseen data. VeriCam works by leveraging the representation power of image models trained for the verification task, where the model develops an intricate feature space that incorporates fine-grained details. By training a model to discriminate between pairs of images from the same and different classes, a relational graph is constructed, representing the class relationships between data points. We then present two approaches for graph clustering: a naive algorithm and a specific setup for the Leiden graph clustering algorithm. The pipeline is validated on the LPLCv2 dataset, which comprises real-world traffic surveillance images. We show that the dataset carries an inherent capture device bias that is posed as a generalization challenge for downstream License Plate recognition tasks such as OCR. As such, we dynamically identify capture devices with a label-agnostic approach, enabling the construction of a fair and unbiased benchmark. In the cross-device scenario, our pipeline reaches an F1-Score of 93.45 in the verification baseline and a V-Measure score of 80.13 in the clustering step. All code is publicly available at https://github.com/lmlwojcik/VeriCam
Read the full abstract
Computer Vision
DreamX-Creator: Democratizing Native Audio-Video Generation at 2K Resolution
Recent video generators often omit audio or synthesize it in a separate stage, limiting reciprocal modeling of visual dynamics and acoustic events. We present DreamX-Creator 1.0, a compact native joint audio-video generation system centered on a 7B generator. Conditioned on a first frame and a text prompt, the generator jointly denoises modality-specialized audio and video streams. The streams are processed independently in the first half of the network and coupled in the latter half through Gated Cross-Modal Attention, whose token- and head-wise output gates modulate each active cross-modal attention-head output. A unified Audio-Video Data System constructs and filters temporally coherent clips, produces structured multimodal annotations, and organizes clips into capability-oriented data pools. Progressive Joint Training comprises two audio-video pre-training stages followed by High-Quality Finetuning. Audio-Video Reinforcement Learning further post-trains the generator with Modality-Aware Multimodal Feedback that routes video-, audio-, and cross-modal feedback to the corresponding streams. For high-resolution output, our Autoregressive 1-Step 2K Refinement pipeline adapts a bidirectional multi-step teacher into an autoregressive multi-step refiner and distills it into a student requiring one denoising evaluation per temporal chunk. Overall, DreamX-Creator 1.0 achieves native, synchronized audio-video generation with performance competitive with state-of-the-art open-source systems. By releasing our compact 7B generator and 2K Refiner, we seek to democratize native audio-video generation and provide an accessible foundation for future research in unified audio-video generative modeling.
Read the full abstract
Artificial Intelligence
BLOOM-WILT: Logit Tilting for Behaviour Elicitation in Automated LLM Auditing
Users of a deployed language model routinely encounter behaviours that have never been observed during testing, since deployment puts the model through orders of magnitude more interactions than any evaluation can simulate. Automated auditors can mitigate this issue as they are flexible enough to discover almost any specified behaviour, yet their random sampling approach makes them inefficient. To address this shortcoming, we introduce BLOOM-WILT, a full auditing pipeline that elicits natural multi-turn instances of rare behaviours, without training cost or access beyond the target's next-token distribution. On the user input side, WILT revises its conversational strategy across rounds, learning from previous scored interactions. On the model output side, WILT generates behaviour-relevant continuations by feeding the model with an elicitation prompt and mixing its next-token distribution with the unprompted model. We evaluate WILT across 4 target models and 8 behaviours, and show that it improves over the baseline in 30 of the 32 settings, thus revealing that a larger number of models exhibit harmful behaviours than previously suggested. Furthermore, WILT is the only method that can raise the mean presence of self-harm encouragement from 51% to 100% in Qwen3.5-4B, without decreasing the output probability of the generated text. Our code is available at https://github.com/AdrSkapars/bloom-wilt.
Read the full abstract
Mathematical Physics
The Dunkl two-photon/Schrödinger algebras and their applications
The work introduces the novel Dunkl two-photon and Dunkl-Schrödinger algebras as reflection-extended counterparts of the standard six-dimensional two-photon and Schrödinger algebras. These extensions are realized through the presence of an involutive generator representing a reflection, and they reduce to the undeformed two-photon and Schrödinger algebras in the absence of this additional generator in an appropriate limit. For both algebraic structures, which are shown to be isomorphic, we derive the associated cubic polynomial Casimir invariant. Furthermore, we identify several structurally significant subalgebras and compute their corresponding quadratic Casimir invariants. Representations of the Dunkl two-photon algebra are constructed on the Fock space of the one-dimensional Dunkl harmonic oscillator and subsequently translated into holomorphic realizations within the associated Dunkl-Fock-Bargmann representation. Finally, for the Dunkl-Schrödinger algebra, we present an explicit vector field realization in terms of the Dunkl spatial derivative, while keeping time continuous. When applied to the Casimir invariants of the aforementioned subalgebras, this realization yields a family of six nontrivial Dunkl differential-difference equations in $1+1$ dimensions, including, in particular, a Dunkl version of the well-known heat-Schrödinger equation.
Read the full abstract
Group Theory
Finitely generated positive cones in $F_n \times \mathbb{Z}$
We construct, for every even $n \ge 2$, a positive cone on $F_n \times \mathbb{Z}$ that is finitely generated as a semigroup, extending the previously known construction for $n = 2$. Malicet, Mann, Rivas and Triestino proved that $F_n \times \mathbb{Z}$ admits an isolated left-order if and only if $n$ is even. Since every finitely generated positive cone determines an isolated left-order, for $n \ge 2$, the group $F_n \times \mathbb{Z}$ admits a finitely generated positive cone if and only if $n$ is even.
Read the full abstract
Software Engineering
LLM Post-Training as Brownfield Maintenance: An Industrial Perspective on Dataware Engineering
Industrial post-training is a brownfield regime. Teams inherit a deployed checkpoint and must land targeted improvements under fixed compute and mixture budgets without regressing the rest. The maintained artifact is increasingly dataware: behavior governed by a curated post-training mixture, updated via bounded mixture patches rather than clean-slate retraining. From an industrial code-generation improvement effort, we offer a maintainer's perspective on why this work is hard in practice, distilling three recurring challenges, zero-sum mixture design, yield as the binding metric, and end-to-end integration under uncertainty, and arguing that progress depends less on one-off recipes than on an engineering discipline for programming dataware. In our case study, interventions that raised the conversion of teacher distillation into usable training data increased accepted supervision by 2.84 times while using the same solution teacher and four solution attempts per candidate problem. In our primary evaluation, the yield-engineered patch improved CodeForces pass@1 by +2.59 points (+3.11 pass@3) and held-out LiveCodeBench v6 pass@1 by +6.11 (+8.05 pass@3), all statistically significant across 16 stochastic evaluations of each benchmark from one fixed checkpoint per condition, with internal AIME and MATH regression suites within tolerance.
Read the full abstract
Functional Analysis
Bang vectors in the real polarization problem: The exact dimensional range
We determine the exact dimensional range in which a normalized longest signed sum, which we call a Bang vector, always satisfies the real polarization inequality: this prescription is universally valid if and only if \(n\leq 14\). For every \(n\geq 15\) we construct \(n\) unit vectors whose all-positive sum is, up to global sign, the unique longest signed sum, while its normalized direction has polarization product smaller than \(n^{-n/2}\). The counterexamples are given by a single explicit family, and for the same configurations we exhibit nonmaximal signed sums whose normalized directions do satisfy the polarization bound.
Read the full abstract
Computation and Language
S3Gym: Can LLMs Turn Self-Testing and Self-Judging into Self-Improvement?
Large language models (LLMs) increasingly interact with external environments and accumulate substantial behavioral experience, yet existing agent benchmarks largely evaluate them as fixed policies. It therefore remains unclear whether an agent can actively test its behavior, judge the resulting experience, and use that experience to improve future decisions. We introduce \textbf{S\textsuperscript{3}Gym}, an interactive benchmark for evaluating LLM self-improvement through three coupled capabilities: Self-Testing, Self-Judging, and Self-Improvement. S$^3$Gym separates permissive exploration from strict held-out evaluation and instantiates this protocol in seven text-based games with executable environment verifiers. We evaluate three pathways for incorporating interaction experience: direct History ICL, score-conditioned Summary Memory, and parameter Training. Our experiments reveal that self-improvement is neither automatic nor uniform. Context-level experience improves performance for several model–game pairs, but the most effective pathway depends strongly on the task structure: summaries are beneficial when experience can be compressed into reusable strategic rules, yet often underperform raw history when success depends on precise, state-contingent information. Parameter training produces substantial gains on some tasks, but also exhibits unstable improvement and severe negative transfer on others. These findings show that recognizing successful actions is insufficient; agents must also transform feedback into executable and transferable policies. S$^3$Gym provides a unified framework for diagnosing this process and identifying the bottlenecks that prevent agents from translating interaction experience into reliable self-improvement.
Read the full abstract
Probability
On (fake) Stationarity in Stochastic Volterra Equations with Affine Drift and Regular Kernels
We investigate the fake stationarity properties of solutions to forward Stochastic Volterra Integral Equations (SVIEs) with affine drift and long-memory (regular) kernels, both on finite horizons and in the long-run regime. By either deriving explicit closed-form specifications for the deterministic initial condition $\phi$ and the mean-reversion function $\mu$ appearing in the drift, or by introducing a deterministic stabilizing factor $\varsigma$ in the diffusion coefficient associated with the kernel while keeping $\mu$ fully flexible, we show that it is possible to induce a fake stationary regime, in the sense that all marginal distributions share the same mean and variance. Afterwards, using a refined asymptotic analysis, we further establish that, in both frameworks, the time-shifted solutions of these long-memory SVIEs converge weakly, in the functional sense, toward a family of $L^2$-stationary processes sharing the same covariance structure, for suitable classes of diffusion coefficients. These results are applied to a class of exponential-fractional Stochastic Volterra Integral Equations driven by an $\alpha$-gamma fractional integration kernel, in the particular regime \(\alpha \geq 1\), which regularizes diffusion paths and invoke textit{ long-term memory}, persistence or long range dependence.
Read the full abstract
Analysis of PDEs
Higher-order Approximate Symmetries
The Fushchych–Shtelen (FS) construction of approximate symmetries, based on a triangular system for the coefficients of a solution expansion, is extended to arbitrary order. A coefficient-map formulation establishes, under explicit hypotheses, a correspondence with the Baikov–Gazizov–Ibragimov (BGI) method. The two approaches are compared for a perturbed cubic wave equation. Nonlinearities admitting an FS point symmetry with prescribed projection onto the unperturbed dilation are classified through second order, up to equivalences. They comprise a power family and logarithmic families at exceptional exponents. Nonlinearities admitting second-order BGI point symmetries that continue this dilation are also classified. Among perturbations nontrivial at first order, these form a proper subfamily of one FS family; another family is admitted by both methods when the perturbation first enters at second order. Joint reduction by the Lorentz algebra and the inherited FS dilation yields explicit invariant solutions for all FS normal forms and recovers the first-order solutions of Fushchych and Shtelen. For the common family with a nontrivial first-order perturbation, the logarithmic corrections arise from expanding a power law with a perturbation-dependent exponent. Periodic traveling waves illustrate the long-scale validity of the FS expansion. The first correction is obtained by quadrature, second-order secular terms are identified, and phase renormalization yields bounded approximations. For integer-power members of the power family, the amplitude dependence of the corrected wavenumber agrees with the symmetry scaling weights.
Read the full abstract
Artificial Intelligence
Cross-Regional Grapevine Cold Hardiness Prediction via Learned Multimodal Latent Representations
Accurate daily predictions of cold hardiness in woody plants are critical in regions where freezing temperatures can damage dormant buds and reduce seasonal yield. Existing biophysical, hybrid, and deep learning models have shown high predictive accuracy when trained on local data but remain largely site-specific. The limited availability of cold hardiness data, coupled with the lack of principled methods for transferring cold hardiness predictions to new regions and cultivars, has limited the broader adoption and practical utility of these approaches, particularly in data-scarce regions. To address these limitations, we propose a cold hardiness prediction framework that learns a transferable latent representation by capturing region-specific variation through learned embeddings. To enable prediction in previously unseen regions, we infer embeddings from (1) text descriptions of the cultivar and growing region, and (2) limited historical observations, supporting both zero-shot and few-shot transfer. Experiments on datasets from six regions across North America demonstrate that our approach consistently outperforms state-of-the-art cold hardiness prediction methods, yielding more accurate predictions and substantially improving transfer to data-scarce regions.
Read the full abstract
Computer Vision
One Adapter, Many Tasks: Task-Conditioned Feature Transformations for Continual Learning
Class-incremental learning (CIL) requires a model to incrementally learn tasks that contain new classes without accessing earlier training data while preserving the ability to recognize all seen classes. Recently, pretrained-model-based approaches have become prevalent by adapting a frozen backbone with additional lightweight trainable modules. Existing methods, however, exhibit limitations: task-specific adapters learn explicit per-task representations but are parameter- and computation-inefficient, while LoRA-based merging methods combine per-task LoRA parameters into a single model whose static aggregated weights cause representation interference during inference. To address these problems, we present FACET: task-conditioned FeAture transformation with ConditionEd feature consisTency, achieving excellent parameter efficiency while producing highly discriminative features during inference. When continually trained on a task sequence, FACET learns a single shared adapter that employs a dynamic task-conditioned feature transformation, shaping the overall feature distribution of the adapter into a mixture of overlap-reduced task-specific components. On the other hand, we propose an efficient replay-free task-conditioned feature consistency loss, aiming to mitigate catastrophic forgetting of the learned mixture distribution in the adapter's feature space. Even when maintaining only a single adapter, FACET demonstrates robust scalability. On both very long task sequences (e.g., 200 tasks) and standard short task sequences (e.g., 20 tasks), our method achieves superior performance while using significantly fewer trainable parameters and GFLOPs. The code will be made open source upon acceptance.
Read the full abstract
Quantum Physics
Tunable Exceptional Points for Quantum Sensing in a Spin–Orbit-Angular-Momentum Coupled BEC
Exceptional points (EPs) can induce strongly amplified responses to weak perturbations, but enhanced spectral sensitivity in conventional non-Hermitian systems does not necessarily translate into improved quantum-limited sensing because of the associated gain and loss noise. Here, we investigate a tunable EP sensing platform based on the intrinsic non-Hermitian Bogoliubov dynamics of a spin-orbit-angular-momentum-coupled Bose-Einstein condensate. Starting from a fully Hermitian microscopic Hamiltonian, we derive a Bogoliubov dynamical matrix that exhibits parity-time (PT) symmetry, with EPs tunable through the Raman coupling and interaction parameters. We identify multiple EPs and map their trajectories and associated stability landscapes, revealing strong quantum Fisher information enhancement when the EPs are approached from the PT-symmetric stable regime. We further find that the gap-opening rate around a second-order EP provides a useful relative indicator for comparing the sensing performance of EPs, while higher-order EPs are not necessarily accessible from the stable regime. Moreover, two second-order EPs can be tuned to overlap, allowing their sensing contributions to add and yielding a linear enhancement of the total quantum Fisher information. Finally, we show that a spin-density measurement can approach the quantum Fisher information limit, providing an experimentally accessible route to tunable EP-enhanced quantum sensing in Bose-Einstein condensates.
Read the full abstract
Computer Vision
Robust retinal identity matching for patient identity verification and retrieval in large ophthalmic databases
Patient identity errors affect the integrity of longitudinal medical imaging records, but retinal photographs contain distinctive anatomical patterns that can be used to match them. We demonstrate, for the first time, that an AI-based retinal identity-matching system trained on a diverse longitudinal dataset is capable of highly accurate identity verification and retrieval across imaging devices, decades of follow-up, and independent cohorts. In the Rotterdam Study, UK Biobank, and Age-Related Eye Disease Study, verification missed only 0.4, 9.3, and 1.1 incorrect assignments per 1,000, respectively, at a fixed rate of five false alerts per 1,000 correct assignments. Retrieval ranked the correct identity first for 99.7%, 97.2%, and 97.6% of queries, respectively, across databases containing thousands of identities. Expert-assisted auditing also uncovered mislabeled images in all three cohorts. Our findings demonstrate the potential of retinal identity matching for reliable, scalable verification and record recovery.
Read the full abstract
Mesoscale and Nanoscale Physics
Device characterization of Si$/$SiGe double quantum dots using exchange oscillations in Earth's magnetic field
Exchange-based semiconductor qubits encompass a broad family of encodings constructed from singlet- and triplet-like spin states, several of which are compatible with operation at zero applied magnetic field. Their reliable operation requires characterization of environmental noise, residual idle interactions, and exchange-dependent decay, but this characterization often relies on multi-axis control calibration or deliberately engineered magnetic-field gradients. A simpler zero-applied-field diagnostic is particularly valuable for hybrid semiconductor-superconductor systems, in which magnetic fields can degrade superconducting components. Here, we use the intrinsic magnetic-field gradient produced by residual nuclear spins in isotopically enriched Si/SiGe to implement exchange oscillations between two quantum dots as a characterization tool without a micromagnet, dynamic nuclear polarization, or prior multi-axis calibration. Using Carr-Purcell-Meiboom-Gill exchange sequences, we extend the singlet coherence from $T_2^*=1.17\pm0.02~\mu$s to $T_2^{\mathrm{CPMG}}=74.8\pm1.8~\mu$s with $N=70$ refocusing pulses. The oscillation phase resolves residual exchange in the tens-of-kilohertz regime and enables it to be mapped across the $(1,1)$ charge cell. These results establish intrinsic-gradient exchange oscillations as a simple, more relevant zero-field diagnostic for exchange-only and related semiconductor qubit encodings that is amenable to rapid, high-throughput device characterization.
Read the full abstract
Optimization and Control
Interpolation Conditions for Instant Data Consistency with Port-Hamiltonian Structure
We develop a data-driven framework for nonlinear port-Hamiltonian (pH) systems based on interpolation conditions to characterize consistency between observed data and structured dynamical models. Specifically, we derive necessary and sufficient conditions for the existence of a pH system with a smooth (convex) Hamiltonian instantly consistent with a given dataset, without requiring explicit parametrization. We further provide a semidefinite programming formulation to verify consistency with non-degenerate interconnection and dissipation structures. Our results provide a principled approach to assess instant data consistency with physical structure and pave the way for control design directly from data.