arXiv:2506.16704v3 Announce Type: replace
Abstract: We study a fundamental question of domain generalization: given a family of domains (i.e., data distributions), how many randomly sampled domains do we need to collect data from in order to learn a model that performs reasonably well on every seen and unseen domain in the family? We model this problem in the PAC framework and introduce a new combinatorial measure, which we call the domain shattering dimension. We show that this dimension characterizes the domain sample complexity. Furthermore, we establish a tight quantitative relationship between the domain shattering dimension and the classic VC dimension, demonstrating that every hypothesis class that is learnable in the standard PAC setting is also learnable in our setting.
Science Journals
arXiv:2510.05566v2 Announce Type: replace-cross
Abstract: Large language models have achieved impressive performance across diverse tasks. However, their tendency to produce overconfident and factually incorrect outputs, known as hallucinations, poses risks in real-world applications. Conformal prediction provides finite-sample, distribution-free coverage guarantees, but standard conformal prediction breaks down under domain shift, often leading to under-coverage and unreliable prediction sets. We propose a new framework called Domain-Shift-Aware Conformal Prediction (DS-CP). Our framework adapts conformal prediction to large language models under domain shift, by systematically reweighting calibration samples based on their proximity to the test prompt, thereby preserving validity while enhancing adaptivity. Our theoretical analysis and experiments on the MMLU benchmark demonstrate that the proposed method delivers more reliable coverage than standard conformal prediction, especially under substantial distribution shifts, while maintaining efficiency. This provides a practical step toward trustworthy uncertainty quantification for large language models in real-world deployment.
arXiv:2510.09288v2 Announce Type: replace-cross
Abstract: The vulnerability of machine learning models to adversarial attacks remains a critical societal security challenge. Traditional defenses, such as adversarial training, typically robustify models by minimizing a worst-case loss. These deterministic approaches do not account for uncertainty in the adversary's attack. While stochastic defenses placing a probability distribution on the adversary exist, they often lack statistical rigor and fail to make explicit their underlying assumptions. To resolve these issues, we introduce a formal Bayesian framework that models adversarial uncertainty through a stochastic channel, articulating all probabilistic assumptions. This yields two robustification strategies: a proactive defense enacted during training, aligned with adversarial training, and a reactive defense enacted during operations, aligned with adversarial purification. Several state-of-the-art defenses can be recovered as limiting cases of our model. We empirically validate our methodology, showcasing the benefits of explicitly modeling adversarial uncertainty.
arXiv:2510.14479v2 Announce Type: replace-cross
Abstract: We consider linear model reduction in both the control and state variables for unconstrained linear-quadratic optimal control problems subject to time-varying parabolic PDEs. The first-order optimality condition for a state-space reduced model naturally leads to a reduced structure of the optimal control. Thus, we consider a control- and state-reduced problem that admits the same minimizer as the solely state-reduced problem. Lower and upper \emph{a posteriori} error bounds for the optimal control and a representation for the error in the optimal function value are provided. These bounds are used in an adaptive algorithm to solve the control problem. We prove its convergence and numerically demonstrate the advantage of combined control and state space reduction.
arXiv:2511.01064v3 Announce Type: replace-cross
Abstract: Variational inference (VI) approximates a target density $p$ by the best match $q$ in a family of tractable distributions. The best variational approximation is found by minimizing a divergence between distributions, $D(p||q)$, and several divergences have been proposed as objective functions for VI, with different choices leading to different approximations. We show that even when these divergences have different minimizers, the resulting approximations all abide by certain symmetry-matching principles. Specifically, our results hold for all $f$-divergences, a broad class which includes the reverse and forward Kullback-Leibler divergences and the $\alpha$-divergences. We show that in the presence of even symmetry, any stationary point of an $f$-divergence is guaranteed to recover the mean of $p$ and likewise, in the presence of elliptical symmetry, any stationary point is guaranteed to recover its correlation matrix. To obtain these guarantees we assume that $p$ and $q$ are unimodal, but notably we do not require them to be log-concave, light-tailed, or even everywhere-smooth. These guarantees generalize a previous result obtained for the reverse Kullback-Leibler divergence when $p$ is log-concave. They also extend to cases where the target density $p$ only exhibits symmetry along some but not all of its coordinates. These partial symmetries arise naturally in Bayesian hierarchical models, where the prior induces a challenging geometry but still possesses axes of symmetry.
arXiv:2606.01926v1 Announce Type: new
Abstract: Generations from large language models often fail to conform to desired constraints such as JSON schema. Existing locally constrained decoding (LCD) approaches enforce constraints by myopically masking out next tokens, resulting in biased sampling and degradation in performance. Recent work uses sequential Monte Carlo (SMC) methods to mitigate such biases, but designing effective proposal distributions or potential functions remains a key challenge. In this work, we propose a generic approach to construct proposals and potentials for SMC sampling from $p_{\mathrm{lm}}( \cdot \mid \mathrm{constraint})$. First, we show that constraints specified as finite automata can be tensorized for efficient execution on GPUs, which we use to construct globally constrained decoding (GCD) proposals. In addition, leveraging the fact that tensorized finite automata share the same circuit structure as hidden Markov models, we circuit-multiply them to obtain the probabilistic GCD (P-GCD) proposals encoding both logical and probabilistic information about the target distributions. We evaluate (P-)GCD on the tasks of function calling, keyword-based generation, and SQL generation. Experiments show that under the same SMC sampling setup, compared to LCD proposals, (P-)GCD converges faster to the target distribution with significantly fewer particles.
arXiv:2606.01421v1 Announce Type: new
Abstract: We show that an array of scatterers which has been designed to have latent ("hidden") symmetries can be used as a sensor. We use the capacitance matrix as a canonical model for three-dimensional hybridisation and study how the introduction of an "intruder'' scatterer breaks the latent symmetries. By analysing the degree to which each symmetry is broken, we identify the radius of the intruder and localize its position. This can be achieved using a dictionary-based approach, however Bayesian inference or an artificial neural network (multi-layer perceptron) perform better in the presence of measurement noise. To our knowledge, this is the first time latent symmetries have been exploited successfully for sensing problems. It is also the first time latent symmetries have been observed in a three-dimensional open system that cannot be approximated by a sparse graph.
arXiv:2603.28894v2 Announce Type: replace-cross
Abstract: We introduce a numerical tensor-network method to compute the statistics of the charge transferred across an interface partitioning an interacting one-dimensional many-body lattice system with $U(1)$ symmetry. Our approach is based on a matrix-product state representation of the process tensor (also known as influence functional or influence matrix) describing the effect of the bulk system on the degrees of freedom at the interface, allowing us to evaluate a multi-time correlation function that yields the moment-generating function of charge transfer. We develop a scheme to truncate non-Markovian correlations which preserves the proper normalization of the process tensor and ensures the correct physical properties of the generating function. We benchmark our approach by simulating magnetization transport within the Heisenberg spin-$1/2$ XXZ brickwork circuit model at infinite temperature. Our results recover the correct transport exponent describing ballistic, superdiffusive, and diffusive transport in different regimes of the model. We also demonstrate anomalous transport encoded by a self-similar scaling form of the moment-generating function outside of the ballistic regime. In particular, we confirm the breakdown of Kardar-Parisi-Zhang universality in higher-order transport cumulants at the isotropic point. Our work paves the way for process-tensor descriptions of non-Markovian open quantum systems to address current fluctuations in strongly interacting systems far from equilibrium.
arXiv:2510.14904v4 Announce Type: replace
Abstract: Dense Video Object Captioning (DVOC) is the task of jointly detecting, tracking, and captioning object trajectories in a video, requiring the ability to understand spatio-temporal details and describe them in natural language. Due to the complexity of the task and the high cost associated with manual annotation, previous approaches resort to training strategies with limited data, potentially leading to suboptimal performance. To circumvent this issue, we propose to generate captions about spatio-temporally localized entities leveraging a state-of-the-art VLM, and extend the LVIS and LV-VIS datasets with our synthetic captions (LVISCap and LV-VISCap). Moreover, we introduce an end-to-end model, CaptionFormer, capable of jointly detecting, segmenting, tracking and captioning object trajectories. CaptionFormer achieves state-of-the-art DVOC results on three existing benchmarks, VidSTG, VLN and BenSMOT. The datasets and code are available at https://www.gabriel.fiastre.fr/captionformer/.
arXiv:2606.02061v1 Announce Type: new
Abstract: Dictionary learning with sparse autoencoders (SAEs) produces overcomplete bases from neural network activations that are often interpretable and reduces polysemanticity. However, features from SAEs vary substantially across random seeds -- a problem known as instability. Archetypal SAEs (Fel et al., 2025) were proposed as a general dictionary-learning intervention for more reliable concept extraction, and report more stable dictionaries at the end of training. We demonstrate that the stability claimed by archetypal SAEs is a result of setting identical initialization across multiple runs. Through our analyses, we attempt to clarify two distinct notions in mechanistic interpretability that may be ambiguously used: stability is agreement between two independently trained models, whereas stabilization is the convergence of independently initialized runs toward a common solution. This distinction is critical for mechanistic interpretability of natural language processing (NLP), where feature stability is increasingly used as evidence that SAE features are reusable units of analysis. Experiments from archetypal SAEs share a deterministic k-means decoder initialization, setting inter-run dictionary distance to zero before training begins. When this initialization is removed, the archetypal constraint provides no stabilization advantage in our setting. We further identify a preprocessing-dependent cosine geometry issue that complicates interpretation of endpoint stability metrics. Overall, our study supports the value of studying SAEs within the larger dictionary-learning tradition while showing that stability claims require trajectory diagnostics and initialization ablations.
arXiv:2603.06989v3 Announce Type: replace
Abstract: This paper introduces MipSLAM, a frequency-aware 3D Gaussian Splatting (3DGS) SLAM framework capable of high-fidelity anti-aliased novel view synthesis and robust pose estimation under varying camera configurations. Existing 3DGS-based SLAM systems often suffer from aliasing artifacts and trajectory drift due to inadequate filtering and purely spatial optimization. To overcome these limitations, we propose an Elliptical Adaptive Anti-aliasing (EAA) algorithm that approximates Gaussian contributions via geometry-aware numerical integration, avoiding costly analytic computation. Furthermore, we present a Spectral-Aware Pose Graph Optimization (SA-PGO) module that reformulates trajectory estimation in the frequency domain, effectively suppressing high-frequency noise and drift through graph Laplacian analysis. Extensive evaluations on Replica and TUM datasets demonstrate that MipSLAM achieves state-of-the-art rendering quality and localization accuracy across multiple resolutions. Code is available at https://github.com/yzli1998/MipSLAM.
arXiv:2605.24634v3 Announce Type: replace
Abstract: Composed image retrieval (CIR) searches a corpus with a reference image and a text describing how to modify it. Despite rapid progress from triplet-trained compositors to zero-shot and generative methods, essentially all systems share one assumption: that a query maps to a single target, scored by Recall@K against one annotation. We argue this is fundamentally at odds with the task. A query such as make it more formal does not name an image but a region of the corpus, and which member the user intends is genuinely underdetermined. This underspecification is the root of the well-known false-negative problem and leaves current models unable to tell a precise query from an ambiguous one. We reframe CIR as calibrated intent resolution under uncertainty: a retriever is wrapped in a conformal prediction layer that returns a candidate set with a coverage guarantee and whose size is a principled measure of ambiguity; when the set is large, an expected-information-gain policy asks the single most useful clarifying question, drawn from interpretable ambiguity axes, and the set contracts. We introduce AmbiCIR, a benchmark and human-validated user simulator that revive the dormant auxiliary and dialogue annotations of CIRR and extend the multiple-positive setting of CIRCO. Across open-domain and fashion benchmarks our method matches single-turn state of the art, confirming calibrated resolution is cost-free on precise queries, while reaching the intended target in a fraction of the interaction budget required by naive conversational baselines, and it is the first to report valid coverage and calibration for the task.
arXiv:2603.07109v2 Announce Type: replace
Abstract: Understanding physical transformations is fundamental for reasoning in dynamic environments. While Vision Language Models (VLMs) show promise in embodied applications, whether they genuinely understand physical transformations remains unclear. We introduce ConservationBench evaluating conservation -- whether physical quantities remain invariant under transformations. Spanning four properties with paired conserving/non-conserving scenarios, we generate and evaluate 23,040 questions across 112 VLMs. Results reveal systematic failure: performance remains near chance with improvements on conservation tasks accompanied by drops on controls. Control experiments show strong textual priors favoring invariance, yet models perform worse with actual visual content when performance is balanced across conserving and non-conserving scenarios. Neither temporal resolution, prompting, nor curated sampling helps. These findings show that current VLMs fail to maintain transformation-invariant representations of physical properties across dynamic scenes.
arXiv:2606.02571v1 Announce Type: new
Abstract: Multilayers of Babinet complementary periodic structures constructed with miniarrays of spherical plasmonic nanoresonators were optimized to ensure Generalized Faraday Rotation. Nonreciprocal rotation and asymmetric transmission were achieved in spectrally overlapping regions due to the reach physics involving (i) symmetry breaking via coupled localized modes, (ii) Brillouin zone-folding stemmed from constituent sub-lattices forming in-plane twisted coupled loops, (iii) interlayer coupling between Babinet complementary patterns. The nanophotonical phenomena include (i) quasi-BIC resonances, (ii) hierarchically coupled localized and propagating modes that results in time-periodic Floquet modulation, (iii) initialization of synthetic potentials tuneable independently via intra and inter-layer parameters. The unique bianisotropic composites result in a synthetic vector gauge and emulated magnetic field manifesting itself in tilted-precessing magnetic dipoles and the accompanying modulation being time-periodic, inherently ensures a synthetic dimension. The asymmetric transmission is enhanced in the classical sense along quantized flat bands, and in mixed and forward bases inside finite wavelength-and-tilting intervals overlapping with nonreciprocal polarization rotation. The transmitted pulse re-shaping proves beating of nearby resonant modes, the loss can be compensated with active ad-layers thereby resulting in Faraday isolator capability. The multilayers synthetize topological phenomena in high-dimensional synthetic parameter spaces.
arXiv:2511.04873v2 Announce Type: replace-cross
Abstract: Prototype selection methods compress a training set, but the existing taxonomy of condensation, edition, hybrid, competence-based, optimization-based, and clustering-based families does not include methods that operate on the multi-scale topological structure of the data. This paper introduces two different persistence-based prototype selector variants, Topological Prototype Selector (TPS) and Boundary-Conscious Topological Prototype Selector (BoundaryTPS). TPS uses two sequential Rips filtrations to retain boundary-relevant and interior-typical points. BoundaryTPS is a single-stage variant whose vertex-weighted filtration concentrates retention near the decision boundary. We evaluate both methods against seven classical baselines on fifteen real datasets and find that the topological methods occupy a different operating point in the prototype-selection design space than existing methods. BoundaryTPS achieves the lowest mean Friedman rank on $H_1$ persistence-diagram preservation and is significantly better than five of the seven baselines (Nemenyi, $\alpha = 0.05$). TPS ranks third on the same endpoint. Both methods are more stable under fold perturbation than any chained-decision selector tested, and both inherit the source set's class proportions without label-aware machinery. On aggregate G-Mean both methods are competitive but not leading, with rank-1 frequencies of $11.3\%$ (TPS) and $9.9\%$ (BoundaryTPS) across fold combinations. Empirically, both methods scale sub-quadratically in sample size.
arXiv:2605.09382v2 Announce Type: replace
Abstract: The Linear Assignment Problem is a fundamental combinatorial optimization task where classical exact solvers ensure optimality but suffer from an $\mathcal{O}(N^{3})$ bottleneck, while recent neural approximations struggle with scalability and exactness. We propose a learning-augmented framework that accelerates exact solvers by predicting dual variables to warm-start the search, backed by a fallback mechanism to preserve worst-case guarantees. Central to our approach is RowDualNet, a lightweight, row-independent architecture that avoids the $\mathcal{O}(N^{2})$ memory bottleneck of graph models, enabling scalable neural warm-starting up to $N=16{,}384$. Feasibility is guaranteed by construction via the Min-Trick mechanism, completely eliminating the need for costly iterative projections. Empirically, our method drastically reduces the search effort of the Jonker-Volgenant (LAPJV) algorithm, yielding robust zero-shot generalization with strict optimality and end-to-end speedups of over 2x on complex synthetic data, 1.25x on real-world tracking, and 1.5x on transportation networks.
arXiv:2606.02196v1 Announce Type: new
Abstract: The rapid expansion of the Python ecosystem has fueled two distinct but converging threats: adversaries increasingly target the software supply chain via the Python Package Index (PyPI), while also building evasive, cross-platform malicious binaries compiled from source code written in Python. Current program analysis techniques struggle to address this dual threat. Static analysis based tools are often blinded by runtime obfuscation and compiled bytecode, while dynamic analysis based ones are fragile, prone to evasion by environmental guardrails, and often terminates prematurely due to unsatisfied dependencies.
To overcome these limitations, we present PyFEX, a resilient forced-execution engine. PyFEX explores a program's behavioral space systematically by forcing execution across all conditional branches to bypass evasion checks. To address the fragility of dynamic execution, it introduces a novel resilient crash recovery mechanism that synthesizes dummy objects to satisfy failed operations at the runtime, allowing analysis to proceed past fatal errors, and employs path merging to mitigate path explosion. PyFEX further incorporates an automated entry identification mechanism that proactively discovers and invokes dormant functions, exposing malicious logic hidden within uncalled APIs. To demonstrate the efficacy of this engine, we built PyFEXScan, a proof-of-concept malware detector built on top of PyFEX. Evaluated against both known malicious PyPI packages and real-world compiled binaries, PyFEX exposes critical behaviors missed by the existing state-of-the-art tools. In a live deployment on PyPI, PyFEXScan discovered 212 previously unknown malicious packages accounting for over 91,648 downloads, underscoring the necessity of resilient, exhaustive analysis for securing the Python ecosystem.
arXiv:2606.00146v1 Announce Type: cross
Abstract: Motion artifacts in magnetic resonance imaging (MRI) degrade diagnostic reliability. Existing deep learning methods are typically contrast-specific and fail to generalize across diverse modalities and artifact severities. We propose a unified framework combining parameter-informed contrast disentanglement with severity-aware adaptive correction. ScanCLIP, pretrained on over 30,000 MRI text-image pairs, derives contrast embeddings from acquisition parameters to disentangle contrast style from anatomical content, yielding contrast-free features. A Vision Transformer then estimates motion severity and routes features through a Mixture-of-Experts network, enabling targeted artifact correction. A dual-pathway decoder reconstructs both the clean image and residual artifact map, enforcing image-space consistency. On IXI and HCP benchmarks, our method improves PSNR by 0.75 dB and SSIM by up to 0.0279 over state-of-the-art approaches, with larger gains at higher artifact severities. It further demonstrates robust zero-shot generalization on real-world clinical data acquired with unseen scanning parameters, where existing methods either fail to remove artifacts or introduce additional distortions.
arXiv:2606.00157v1 Announce Type: cross
Abstract: We consider establishing the interpretability theory of deep learning through constructing a corresponding relationship between the renormalization group (RG) method in statistical physics and the training process of deep neural networks (DNNs). We have proved the constructed relationship using the one-dimensional Ising model as the input data. In this paper we generalize our results to the case of continuous input data, which is a necessary preparation for applying the corresponding framework to real-world data. To be representative, we consider a class of data distribution in the exponential family. We prove that when the parameters of fully connected (FC) DNNs achieve their optimal value after training, the characteristic parameters of the feature layer output of DNNs are equal to the fixed points of the characteristic parameters of input data under RG method for continuous fields. This conclusion shows that the training process of DNNs is equivalent to RG calculation on this kind of data and therefore the network can extract main features from the input data just like RG. Also, the equivalence further validates the correspondence framework we have established, providing an explanation for the outstanding performance of DNNs on real-world data.
arXiv:2606.02241v1 Announce Type: new
Abstract: Is the uniform-state diffusion framework a more powerful paradigm for discrete diffusion? Recent studies indicate that this may be the case. In combination with predictor-corrector samplers, uniform-state diffusion models (USDMs) produce samples of higher-quality than masked diffusion models (MDMs), and USDMs equal or outperform MDMs in downstream tasks, even though they exhibit greater perplexity. Two issues remain unresolved. First, existing work compares uniform and masked diffusion with un-informed correctors that re-inject noise at random positions, rather than targeting tokens most likely to be wrong. Second, prior work compares full-sequence diffusion models, so we do not know whether the same conclusion holds when tokens are generated block by block. To address these issues, we introduce BlockGen, a blockwise sequence model that we instantiate with both masked and uniform diffusion. BlockGen trains on a mixture of block sizes and its likelihood interpolates between AR and pure diffusion more finely than models with a fixed block size. BlockGen enables AR-informed predictor-corrector sampling (ARPC), which combines AR and diffusion predictions to re-generate unlikely tokens without an auxiliary verifier. Under ancestral sampling, uniform outperforms masked in the block-by-block setting, especially in the few-step regime. Under ARPC, the gap closes and reverses at high NFE. With block size $16$ on GSM8K, MDMs reach slightly higher accuracy than USDMs, and we observe a similar trend in Generative Perplexity on OpenWebText. Find our code at https://github.com/jdeschena/blockgen.
arXiv:2606.00473v1 Announce Type: cross
Abstract: Ionizing radiation in the form of $\alpha$, $\beta$, $\gamma$, and additional high-energy particles can induce decoherence via phonon and quasiparticle poisoning in superconducting qubits. Recent studies have explored this effect using cosmic rays or controlled radioactive sources held in the proximity of a qubit package, and have concluded that reductions in such ``external'' environmental radiation may benefit stable operation of qubit devices. However, the effect of long-lived, unstable daughters of $^{222}$Rn that ``plate out'' directly on device and packaging surfaces has not been as extensively explored. This plate-out process, well-known to the dark matter direct detection field, occurs throughout the fabrication and testing lifecycle of a device and (separately) its packaging, and produces a local source of $\alpha$-decays which can remain active for decades. As this scales with chip area, understanding and managing this source of ionizing radiation is relevant for successfully scaling quantum computing architectures to larger numbers of qubits in a radiation-robust way. We present a setup capable of accelerating and enhancing radon daughter plateout by a factor of $7\times10^4$ over ambient, in order to study, \textit{in situ}, the impact of these events on superconducting qubits. We also provide outlook on the potential impact of this source of ionizing radiation on current and future qubit arrays.
Solar Daylighting to Offset LED Lighting in Vertical Farming: A Techno-Economic Study of Light Pipes
arXiv:2603.15806v2 Announce Type: replace
Abstract: Vertical farming is a controlled-environment agriculture (CEA) approach in which crops are grown in stacked layers under regulated climate and lighting, enabling predictable production but requiring high electricity input. This study quantifies the techno-economic impact of roof-mounted daylighting in a three-tier container vertical farm using a light-pipe (LP) system that delivers sunlight to the upper tier. The optical chain, comprising a straight duct and a tilting aluminum-coated mirror within a rotating dome, was modelled in Tonatiuh to estimate crop-level photon delivery and solar gains. These outputs were coupled with a transient AGRI-Energy model to perform year-round simulations for Dubai. Tier-3 strategies were compared against a fully LED benchmark, including daylight-only operation, on/off supplementation, PWM dimming, UV-IR filtering, variable-transmittance control, and simple glazing. Ray-tracing predicted an overall LP optical efficiency of 45%-75%, depending on solar position, quantifying the fraction of incident daylight at the collector aperture delivered to the target growing zone. Daylight-only operation reduced the total three-tier yield by 17% and was not economically viable despite 27-29% electricity savings. Hybrid daylight-LED strategies preserved benchmark yield while reducing electricity use. PWM dimming combined with UV-IR filtering achieved the lowest specific electricity energy consumption (6.32 kWh/kg), 14% below the benchmark. Overall, viability remains CAPEX-limited because achievable electricity savings are insufficient to offset the added investment and thus improves mainly under high electricity and carbon-price contexts, although the LP system delivers a 15-38% lower light cost than an optical-fiber reference under identical incident daylight.
arXiv:2606.01513v1 Announce Type: new
Abstract: High-stakes enterprise document generation, including financial dispute narratives, compliance notices, and audit summaries, demands schema correctness, policy compliance, and low-latency operation at scale. Prior to a unified guardrail layer, production systems often stitched together separate PII redaction, content moderation, and format validation steps, leading to fragmented logic, slower request paths, and higher operational cost.
We present a guardrail orchestration layer for text and image inputs that couples multi-candidate generation with an explicit compliance score used for early exit. The framework runs configurable parallel generation heads, scores candidates against weighted guardrails including PII detection, content moderation, schema constraints, and domain rules, and returns the best-scoring output with selection metadata. The available operational readout reports 5 attempts within 20 seconds and 91 percent compliance.
For payments dispute defense summaries, we analyze aggregate operational scenario readouts rather than a randomized A/B test. Variable cohorts show higher count win rates than controls overall, 301/659 versus 536/1548, corresponding to +11.0 percentage points with 95 percent confidence interval [6.6, 15.5] and p < 0.001, and for adjusted item-not-received cases, +7.5 percentage points with 95 percent confidence interval [0.2, 15.7] and p = 0.045. Fraud and local evidence-ranking deltas are directionally positive but not statistically significant from the aggregate count data. We also report reviewer-calibrated Responsible-AI evidence-quality signals from 770 generated-evidence reviews and a 70-case OCR slice, and document the reproducibility boundary through the request interface, scoring logic, pseudocode, and operational evidence boundary.
arXiv:2512.10958v2 Announce Type: replace
Abstract: Generative world models are reshaping embodied AI, enabling agents to synthesize realistic 4D driving environments that look convincing but often fail physically or behaviorally. Despite rapid progress, the field still lacks a unified way to assess whether generated worlds preserve geometry, obey physics, or support reliable control. We introduce WorldLens, a full-spectrum benchmark evaluating how well a model builds, understands, and behaves within its generated world. It spans five aspects -- Generation, Reconstruction, Action-Following, Downstream Task, and Human Preference -- jointly covering visual realism, geometric consistency, physical plausibility, and functional reliability. Across these dimensions, no existing world model excels universally: those with strong textures often violate physics, while geometry-stable ones lack behavioral fidelity. To align objective metrics with human judgment, we further construct WorldLens-26K, a large-scale dataset of human-annotated videos with numerical scores and textual rationales, and develop WorldLens-Agent, an evaluation model distilled from these annotations to enable scalable, explainable scoring. Together, the benchmark, dataset, and agent form a unified ecosystem for measuring world fidelity -- standardizing how future models are judged not only by how real they look, but by how real they behave.
arXiv:2512.12634v4 Announce Type: replace
Abstract: Mobile GUI Agents, AI agents capable of interacting with mobile applications on behalf of users, have the potential to transform human computer interaction. However, current evaluation practices for GUI agents face two fundamental limitations. First, they either rely on single path offline benchmarks or online live benchmarks. Offline benchmarks using static, single path annotated datasets unfairly penalize valid alternative actions, while online benchmarks suffer from poor scalability and reproducibility due to the dynamic and unpredictable nature of live evaluation. Second, existing benchmarks treat agents as monolithic black boxes, overlooking the contributions of individual components, which often leads to unfair comparisons or obscures key performance bottlenecks. To address these limitations, we present MobiBench, the first modular and multi path aware offline benchmarking framework for mobile GUI agents that enables high fidelity, scalable, and reproducible evaluation entirely in offline settings. Our experiments demonstrate that MobiBench achieves 94.72 percent agreement with human evaluators, on par with carefully engineered online benchmarks, while preserving the scalability and reproducibility of static offline benchmarks. Furthermore, our comprehensive module level analysis uncovers several key insights, including a systematic evaluation of diverse techniques used in mobile GUI agents, optimal module configurations across model scales, the inherent limitations of current LFMs, and actionable guidelines for designing more capable and cost efficient mobile agents.