arXiv:2510.11917v3 Announce Type: replace
Abstract: Dementia disorders such as Alzheimer's disease (AD) and frontotemporal dementia (FTD) exhibit overlapping electrophysiological signatures in EEG that challenge accurate diagnosis. Existing EEG-based methods are limited by full-band frequency analysis, which hinders the precise differentiation of dementia subtypes and severity stages. To address this limitation, we propose a Variational Mixture of Graph Neural Experts (VMoGE) framework that integrates multi-band EEG analysis with variational graph neural networks and a mixture-of-experts architecture. Each expert specializes in a specific EEG frequency band and models brain connectivity using a Gaussian Markov random field prior, while a variational gating mechanism adaptively integrates the expert outputs. This design enables the model to learn frequency-specific brain network representations while modeling latent uncertainty through variational inference. Experimental results on two EEG dementia datasets show that VMoGE achieves strong performance, with an AUC of 0.89 for HC vs. AD classification in the main comparison and competitive results across dementia subtyping and CDR staging tasks. Clinically, VMoGE offers three key translational values: the expert gating weights correlate with MMSE scores and CDR severity; slow-wave $\delta$- and $\theta$-band contributions are associated with AD-related EEG slowing and disease progression; and spatially localized activation maps reveal posterior $\theta$- and $\alpha$-band alterations and region-specific $\beta$-band changes, providing neurophysiologically interpretable markers aligned with known AD neuropathology.
Science Journals
arXiv:2510.05860v2 Announce Type: replace
Abstract: Policymakers enact and revise privacy laws expecting meaningful benefits for their people in practice. While scholarship has measured the real-world impact of some privacy regulations-the EU and California most notably-limited empirical evidence exists for many of the more than 140 countries that have implemented some form of privacy legislation. Switzerland, a multilingual country bordered almost entirely by EU states, is one such example.
This paper analyzes the extent to which a 2023 alignment of Swiss privacy law with EU privacy regulation affected website privacy policies in Switzerland. To address Switzerland's unique multilingual culture, we develop an LLM-based pipeline that extracts legally relevant information as document-level labels in a single inference without requiring translation. On a benchmark of 120 expert-annotated privacy policies in German, French, Italian, and English, our pipeline achieves F1 scores above 0.90 for most pairs of languages and legally relevant disclosures.
Applying this pipeline to privacy policies we collected from more than 35,000 Swiss- and EU-facing websites before and after the 2023 privacy law revision, we find significant increases in both mandatory and voluntary disclosures of data subject rights among Swiss privacy policies. In exploring the mechanisms driving increased disclosure rates, we discover heavy use of automated privacy policy generators and find that generated policies are associated with up to 15 percentage points higher disclosure rates. These results provide large-scale empirical evidence of how regulatory change and novel drafting technologies impact the content of privacy policies in a unique multilingual environment.
arXiv:2510.12363v4 Announce Type: replace
Abstract: The pretraining-finetuning paradigm has facilitated numerous transformative advancements in artificial intelligence research in recent years. However, in the domain of reinforcement learning (RL) for robot locomotion, individual skills are often learned from scratch despite the high likelihood that some generalizable knowledge is shared across all task-specific policies belonging to the same robot embodiment. This work aims to define a paradigm for pretraining neural network models that encapsulate such knowledge and can subsequently serve as a basis for warm-starting the RL process in classic actor-critic algorithms, such as Proximal Policy Optimization (PPO). We begin with a task-agnostic exploration-based data collection algorithm to gather diverse, dynamic transition data, which is then used to train a Proprioceptive Inverse Dynamics Model (PIDM) through supervised learning. The pretrained weights are then loaded into both the actor and critic networks to warm-start the policy optimization of actual tasks. We systematically validated our proposed method with 9 distinct robot locomotion RL environments comprising 3 different robot embodiments, showing significant benefits of this initialization strategy. Our proposed approach on average improves sample efficiency by 36.2% and task performance by 4.3% compared to random initialization. We further present key ablation studies and empirical analyses that shed light on the mechanisms behind the effectiveness of this method.
arXiv:2607.13633v1 Announce Type: new
Abstract: Prompt extraction of victims from water is crucial in water surface rescue missions. However, previous research on rescue robots has seldom addressed this issue. This paper presents QuadBoat, a bio-inspired unmanned surface vehicle (USV) designed to track and retrieve victims from water. QuadBoat features a quadrupedal robot configuration, enabling it with highly adaptable and agile maneuverability through its actively adjustable posture. Employing an inverse kinematics-based controller and cascaded model predictive control (MPC)-PID controller for overall movement, QuadBoat can accurately track and retrieve objects on the water surface. Maneuverability demonstrations validate QuadBoat's high agility, while a series of tracking experiments, including leg action tracking and trajectory tracking, confirm its high motion accuracy and system mobility. Finally, visual-based tracking and object pickup experiments further verify QuadBoat's target tracking capabilities and its effectiveness in executing rescues, both indoors and outdoors.
arXiv:2607.13693v1 Announce Type: new
Abstract: This book chapter covers the evolution of social simulation from classical agent-based models, in which agents interact according to explicitly defined behavioral rules, to AI-enhanced simulations based on Large Language Models and, ultimately, Social Digital Twins: high-fidelity, data-driven representations of real-world socio-technical systems. Along this trajectory, we discuss the main methodological foundations, applications, advantages, and limitations of each paradigm, highlighting the progressive shift from abstract models designed to investigate general social mechanisms toward increasingly realistic computational representations of specific social systems.
arXiv:2607.14088v1 Announce Type: new
Abstract: Video generative models commonly rely on latent spaces learned by 3D Variational Autoencoders (3D-VAEs). However, conventional 3D-VAEs are mainly optimized for pixel-level reconstruction, which can limit the semantic and spatio-temporal structure captured by their latents. Meanwhile, Video Foundation Models (VFMs) such as V-JEPA 2 and VideoMAEv2 show strong video understanding capabilities, yet whether their frozen representations can be transformed into compact, reconstruction-capable, and generation-friendly video latents remains largely unexplored. We answer this question with VideoRAE, a representation autoencoder that leverages multi-scale hierarchical features from a frozen video foundation encoder and compresses them with a lightweight 1D self-attention projector. VideoRAE supports both continuous latents for Diffusion Transformers and discrete tokens for autoregressive models via multi-codebook high-dimensional quantization. During decoding, a local-and-global representation alignment objective with the frozen VFM teacher improves semantic preservation and enables training without KL regularization. Experiments show that VideoRAE achieves strong reconstruction in both continuous and discrete regimes. On UCF-101, it obtains state-of-the-art class-to-video gFVDs of 40 and 93 with AR and DiT generators, respectively, while converging approximately 5x faster than competing autoencoder baselines. In a controlled 2B-scale text-to-video study, replacing LTX-VAE with VideoRAE leads to faster convergence under comparable settings. These results validate frozen VFM representations as versatile and generation-friendly video latents. The model and code will be released on https://zhxie0117.github.io/VideoRAE.
arXiv:2607.13493v1 Announce Type: new
Abstract: Incremental video search requires high-quality ranking after each keystroke, where intent is often underspecified (e.g., 1-3 character prefixes). We present a personalization system for Apple TV search that combines complementary semantic and collaborative signals at ranking time. Our approach learns two item embedding spaces: (i) a text-based multilingual encoder (TextEmb) fine-tuned on co-engagement triplets via contrastive learning, and (ii) an ID-based collaborative embedding model (IdEmb) trained on interaction-derived positives. At serving time, we construct user representations from recent watch history and inject text- and ID-based user-item cosine similarities into a pairwise XGBoost ranker.
We evaluate with temporally held-out offline datasets and a three-week online controlled experiment. Offline, for sessions with user history, the personalized ranker improves NDCG@10 by 2.99% and MRR by 3.30% over the non-personalized baseline. Slice analyses show that personalization is most needed in incremental search, where intent is still forming: on ambiguous prefix queries (1-3 characters), NDCG@10 lift is +8.63%, versus +1.46% on longer, fully specified queries. Longer-history users benefit more: NDCG lift rises from +2.13% for users with 1-5 history items to +4.37% for users with 51-100, even though baseline relevance is lower for these cohorts (NDCG@10 drops from 0.733 to 0.680), indicating that personalization adds the most value where default ranking underperforms. Online, treatment yields statistically significant gains of +1.14% tap-through rate and +1.23% conversion rate, with a 2.91% improvement in converted-item rank position. We further analyze coverage-precision trade-offs between semantic and collaborative embeddings via ablations isolating each signal, and evaluate embedding quality on a held-out corpus with LLM-judged similarity labels to reduce click/exposure bias.
arXiv:2602.09616v2 Announce Type: replace
Abstract: Reliable retrieval-augmented generation (RAG) systems depend fundamentally on the retriever's ability to find relevant information. We show that neural retrievers used in RAG systems have blind spots, which we define as the failure to retrieve entities that are relevant to the query, but have low similarity to the query embedding. We investigate the training-induced biases that cause such blind spot entities to be mapped to inaccessible parts of the embedding space, resulting in low retrievability. Using a large-scale dataset constructed from Wikidata relations and first paragraphs of Wikipedia, and our proposed Retrieval Probability Score (RPS), we show that blind spot risk in standard retrievers (e.g., CONTRIEVER, REASONIR) can be predicted pre-index from entity embedding geometry, avoiding expensive retrieval evaluations. To address these blind spots, we introduce ARGUS, a pipeline that enables the retrievability of high-risk (low-RPS) entities through targeted document augmentation from a knowledge base (KB), first paragraphs of Wikipedia, in our case. Extensive experiments on BRIGHT, IMPLIRET, and RAR-B show that ARGUS achieves consistent improvements across all evaluated retrievers (averaging +3.4 nDCG@5 and +4.5 nDCG@10 absolute points), with substantially larger gains in challenging subsets. These results establish that preemptively remedying blind spots is critical for building robust and trustworthy RAG systems.
arXiv:2602.09625v3 Announce Type: replace
Abstract: We study the strong-field interaction of a helical bi-chromatic pump with an anisotropic and inhomogeneous molecular system in the form of planar distribution of radially aligned molecular ensemble. This setting gives rise to macroscopic spin-orbit interaction where High Harmonic radiation is emitted while imbued with Orbital Angular Momentum (OAM) whose sign is directly dictated by the helicity of the pump field. We demonstrate this phenomenon in ensembles of $H_2^+$ and $N_2$ molecules with Time-Dependent Density Functional Theory (TDDFT) simulations.
arXiv:2607.13617v1 Announce Type: new
Abstract: Durable workflow engines reconstruct execution state by deterministically replaying an immutable event log, coupling every in-flight run to the code version that produced its history: a new deployment can invalidate the replay of runs started under the old version, silently corrupting state or halting progress. Existing mitigations -- pinning, patch gates, side-by-side deployment -- treat every change as maximally dangerous and drain old versions, untenable for workflows that sleep for weeks.
We present a closed-form probabilistic model that quantifies the risk of upgrading in-flight runs from workflow version $V_1$ to $V_2$ using only a static structural diff and telemetry the protocol already persists -- event logs, step payloads, historical paths -- with no dry-run, sandbox, or shadow execution. Risk decomposes along three axes (protocol, interface, state migration) and combines an exact backward (rehydration) term, computed on recorded prefixes modulo trace equivalence of concurrent completions, with a probabilistic forward term from hitting probabilities in an empirically estimated Markov model of control flow. Estimation is Bayesian throughout, so the Workflow Upgrade Risk (WUR) score carries a credible interval and thin telemetry surfaces as uncertainty. We prove that a zero backward-risk verdict certifies safe rehydration under the new version, and derive a policy partitioning runs into migrate, review, and pin classes. Finally we drop the inter-run independence assumption: coupling through hooks, hierarchy, and shared resources is captured by an empirical coupling graph, fleet risk becomes the least fixpoint of a failure-contagion operator, and the coupling-aware migrate/pin partition is computed exactly as a minimum s-t cut.
arXiv:2510.04876v3 Announce Type: replace
Abstract: Benthic habitat mapping is fundamental for understanding marine ecosystems, guiding conservation efforts, and supporting sustainable resource management. Yet, the scarcity of large, annotated datasets limits the development and benchmarking of machine learning models in this domain. This paper introduces a thorough multi-modal dataset, comprising about a million side-scan sonar (SSS) tiles collected along the coast of Catalonia (Spain), complemented by bathymetric maps and a set of co-registered optical images from targeted surveys using an autonomous underwater vehicle (AUV). Approximately 36000 of the SSS tiles have been manually annotated with segmentation masks to enable supervised fine-tuning of classification models. All the raw sensor data, together with mosaics, are also released to support further exploration and algorithm development. To address challenges in multi-sensor data fusion for AUVs, we spatially associate optical images with corresponding SSS tiles, facilitating self-supervised, cross-modal representation learning. Accompanying open-source preprocessing and annotation tools are provided to enhance accessibility and encourage research. This resource aims to establish a standardized benchmark for underwater habitat mapping, promoting advancements in autonomous seafloor classification and multi-sensor integration.
When Agents Disagree With Themselves: Behavioral Consistency as an Uncertainty Signal for LLM Agents
arXiv:2602.11619v2 Announce Type: replace
Abstract: Running the same LLM agent on identical inputs yields 2.3-4.2 distinct action sequences per 10 runs; this behavioral variance constitutes a training-free, black-box uncertainty signal that instantiates selective classification and distribution-free calibration for agentic systems. Across 8,000 runs of four models on 200 HotpotQA questions, consistent tasks (at most 2 unique paths) achieve 82-87% accuracy while inconsistent tasks (4 or more paths) achieve 41-65%, a gap that survives controls for task difficulty. Divergence concentrates at step 2 (50.5% of Llama tasks), and consistency metrics detect failures with AUROC 0.62-0.78. Exploiting this signal, selective prediction (answering only when k=3 runs agree) achieves 87-88% accuracy at 54-62% coverage, a 6-14pp gain over single-run baselines, and matches a split-conformal baseline without a held-out calibration set. A cross-benchmark validation on SWE-bench (50 tasks, 1,000 runs) preserves the consistency hierarchy while revealing an ~8x spread in mean trajectory length across models, and bootstrap analysis shows single-run evaluations misrank models 29.3% of the time.
arXiv:2510.11686v2 Announce Type: replace
Abstract: Reinforcement learning (RL) promises to expand the capabilities of language models, but it is unclear if current RL techniques promote the discovery of novel behaviors, or simply sharpen those already present in the base model. In this paper, we investigate the value of deliberate exploration -- explicitly incentivizing the model to discover novel and diverse behaviors -- and aim to understand how the knowledge in pre-trained models can guide this search. Our main finding is that exploration with a simple, principled, representation-based bonus derived from the pre-trained language model's hidden states significantly improves diversity and pass@k rates -- both for post-training, and in a novel inference-time scaling setting we introduce. For inference-time, exploration with representation-based diversity improves efficiency, consistently improving pass@k rates across a variety of models and reasoning tasks. For example, for Qwen-2.5-14b-Instruct we obtain over 50% improvement in verifier efficiency on almost all tasks. For post-training, we show that integrating this exploration strategy into an RL pipeline improves reasoning performance over that of the initial model and over standard RL post-training. For example, on AIME 2024, our post-trained Qwen-2.5-7b-Instruct's pass@80 matches the pass@256 of GRPO on the same model, demonstrating a 3x improvement in test-time sample efficiency. Overall, our findings suggest that deliberate exploration -- with the right notion of diversity -- is a practical path toward discovery of new behaviors beyond sharpening.
arXiv:2607.14026v1 Announce Type: new
Abstract: Translating next-month housing-guarantee risk forecasts into auditable operational reports is essential yet challenging because upper-tail events are sparse, source records are confidential, and generated narratives can distort the underlying evidence. Using monthly South Korean \textit{jeonse} deposit guarantee data from September 2015 to December 2025, we introduce an evidence-constrained reporting pipeline that prioritizes upper-tail monitoring, retrieves historical precedents aligned with the forecasting rationale, organizes admissible information into typed evidence contracts, and verifies generated claims before analyst review. We train and select the forecasting backbone on the original panel, whereas the reporting experiments use synthetic aggregate scenarios calibrated to its empirical ranges and temporal structure. The selected forecasting model substantially improves high-risk detection while retaining competitive average error. Across eight LLMs, structured evidence consistently increases report quality, numerical fidelity, and claim-level grounding. A practitioner evaluation involving 51 analysts and related domain professionals further indicates that the reports support real-world review and decision-making: most participants rated them as practically useful and endorsed an operational pilot. These findings demonstrate that reliable LLM-assisted reporting requires predictive models to be coupled with structured evidence, explicit verification, and analyst oversight.
arXiv:2607.13351v1 Announce Type: cross
Abstract: Fault-tolerant quantum computing requires a continuous hybrid quantum error correction (QEC) pipeline comprising measurement readout, syndrome transport, decoding, feedback, and control. Existing QEC simulators primarily evaluate circuits, noise models, decoders, and protocol-level outcomes. System architects, however, must also understand how these workloads contend for and queue across controller, compute, accelerator, and communication resources during protected logical execution. We introduce StreamingQEC, a system-level simulator that translates fault-tolerant logical workloads into resource-constrained streaming-QEC pipelines. An explicit discrete-event simulation provides the reference execution semantics. An automatic staged-fluid mode enables faster approximate design-space exploration, while a certified recurrence mechanism compresses repeated transitions only when their scheduling state and metric contributions match those of the explicit execution trace. We assemble a decoder-runtime dataset containing 9,998 measurements, of which 8,174 are used to fit performance profiles. Recurrence reproduces the reported explicit-simulation metrics across 35 calibrated-profile configurations, as well as additional workload and cadence validation cases. For a 16-job anchor workload, it preserves 59,743,936 decoding events while achieving a 24.0x host-side speedup, and recurrent simulations scale beyond 1.22 billion events. Across 17 reference configurations, the automatics taged-fluid mode yields a mean makespan error of 2.60% and a worst-case error of 6.45%. Design-space studies reveal transfer-limited resource matching,decoder-driven pipeline stalls, and saturation of dedicated resources under microsecond-scale QEC cycles.
arXiv:2602.12811v2 Announce Type: replace
Abstract: When humans and large language models (LLMs) process the same text, activations in the LLMs correlate with brain activity measured, e.g., with functional magnetic resonance imaging (fMRI). Moreover, it has been shown that, as the training of an LLM progresses, the performance in predicting brain activity from its internal activations improves more in the left hemisphere than in the right one. The aim of the present work is to understand which kind of competence acquired by the LLMs underlies the emergence of this left-right asymmetry. Using the OLMo-2 7B language model at various training checkpoints and fMRI data from English participants, we compare the evolution of the left-right asymmetry in the correlation between brain activity and model predictions alongside performance on several benchmarks. We observe that the asymmetry co-emerges with the formal linguistic abilities of the LLM. These abilities are demonstrated in two ways: by the model's capacity to assign a higher probability to an acceptable sentence than to a grammatically unacceptable one within a minimal contrasting pair, and by its ability to produce well-formed text. By contrast, the left-right asymmetry does not align with the performance on arithmetic or Dyck language tasks; nor with text-based tasks involving world knowledge and reasoning. We generalize these results to another family of LLMs (Pythia) and two other languages, French and Chinese. Our observations indicate that the left-right asymmetry in brain predictivity matches the progress in formal linguistic competence.
arXiv:2602.14091v2 Announce Type: replace
Abstract: In Japan, severe rice shortages in 2024 sparked widespread public controversy across both news media and social platforms, culminating in what has been termed the "Reiwa Rice Riot." This study proposes a framework to analyze the temporal dynamics and causal interactions of emotions expressed on X (formerly Twitter) and in news articles, using the "Reiwa Rice Riot" as a case study. While recent studies have shown that emotions mutually influence each other between social and mass media, the patterns and transmission pathways of such emotional shifts remain insufficiently understood. To address this gap, we applied a machine learning-based emotion classification grounded in Plutchik's eight basic emotions to analyze posts from X and domestic news articles. Our findings reveal that emotional shifts and information dissemination on X preceded those in news media. Furthermore, in both media platforms, the fear was initially the most dominant emotion, but over time intersected with hope which ultimately became the prevailing emotion. Our findings suggest that patterns in emotional expressions on social media may serve as a lens for exploring broader social dynamics.
arXiv:2607.13698v1 Announce Type: new
Abstract: In recent years, there has been growing interest in robust robotic systems for precise bin-picking applications. To achieve reliable performance, such systems must address errors arising from both the object pose estimation and the grasping process. Although various approaches have been proposed, they typically target specific challenges and do not offer general solutions. In this paper, we present a modular framework that jointly handles both error types. The framework incorporates object pose distribution estimation to account for pose uncertainty, which frequently arises in situations with ambiguous observations where a single correct pose cannot be determined. To further reduce uncertainty, we introduce a second-viewpoint module that computes complementary pose distributions, which are subsequently fused. This fusion decreases overall uncertainty and improves system efficiency. Additionally, two independent modules are included to compensate for grasping errors. The modular design allows the components to be combined for optimal performance or used individually, depending on the physical setup.
The proposed method is evaluated in a real-world setup with three different objects, with no errors, and all modules are shown to improve efficiency. These results suggest that incorporating pose distributions with grasping pose errors is a promising direction for developing more flexible and reliable robotic production systems.
To the best of our knowledge, this is the first framework that jointly addresses both grasping and object pose uncertainties using interchangeable modules. We believe there is ample opportunity to integrate additional modules, resulting in improved performance and flexibility. The current framework is limited to pose uncertainties in SO(2), but it could be extended to SE(3), enabling additional modules to improve the system.
arXiv:2607.13789v1 Announce Type: new
Abstract: This paper introduces the Dynamical Vehicle Orienteering Problem (DVOP), a generalization of the Orienteering Problem (OP). The OP maximizes the reward collected from spatial targets under a limited travel budget; the DVOP extends it by accounting for both external and vehicle-actuated forces. We study the DVOP in the context of multi-rotor Unmanned Aerial Vehicle (UAV) flight planning, using a three-dimensional Point-Mass Model (PMM) constrained by maximum velocity and acceleration magnitudes and subject to gravitational acceleration, with the travel budget expressed as a maximum flight time. Because the DVOP couples reward maximization with time-optimal trajectory planning, it cannot be formulated as a simple graph problem and solved exactly without relaxing or under-actuating the vehicle dynamics. We therefore propose two solution approaches: a Branch-and-Bound (BnB) procedure that combines Non-Linear Programming (NLP) and Mixed-Integer Linear Programming (MILP) to provide high-quality solutions, and a Large Neighborhood Search (LNS) metaheuristic that supplies an initial reward bound and scales to instances intractable for the BnB. The BnB relies on a novel MILP formulation of travel costs based on minimum-time trajectory primitives through target triplets, yielding a tight reward upper bound, while the LNS uses limited thrust decomposition to compute fast, high-quality PMM trajectories. Experiments on benchmark instances show improvements of up to 37 % over state-of-the-art solutions for the Kinematic Orienteering Problem, and a real-world deployment on a multi-rotor UAV verifies the proposed PMM solution trajectories.
arXiv:2607.14046v1 Announce Type: new
Abstract: This paper presents Earthquaker-AI, a hybrid educational framework building upon a previously implemented educational robotics project by integrating a conversational AI assistant based on Retrieval-Augmented Generation. It aims to enhance earthquake preparedness and conscious action among primary-school students. The system extends the award-winning STEM project Earthquaker moving from mechanical simulation with Lego WeDo2 to cognitive and metacognitive processing. The robotics component uses Lego WeDo2 automation to simulate seismic response, letting students interact with sensors and actuators as tangible representations of protective actions. The assistant operates as a guided learning mechanism aligning student responses with safety guidelines, while providing rubric-based verbal feedback that supports self-regulated learning and calmness under emergency conditions. Earthquaker-AI follows a progressive learning trajectory aligned with cognitive development. In early grades, the focus is on basic recognition of safety actions through multiple-choice questions, assessed via a two-dimensional rubric. In middle grades, students identify correct action sequences through multiple-choice questions, evaluated via a three-axis rubric. In upper grades, the approach shifts to verbal production, requiring short written responses assessed via a four-dimensional rubric that includes clarity of expression. The dialogic module uses RAG to match student queries semantically with official guidelines, generating safe, accurate responses. Experimental evaluation shows high groundedness and accuracy, with a low hallucination rate. Overall, Earthquaker-AI combines hands-on engagement, information processing, and reflective practice. Combining robotics, rubrics, and AI promotes technological literacy, self-regulation, and responsible use of digital systems, contributing to early crisis-management skills.
arXiv:2603.10330v2 Announce Type: replace
Abstract: Autonomous driving in complex traffic requires planners that generalize beyond hand-crafted rules, motivating data-driven approaches that learn behavior from expert demonstrations. Diffusion-based trajectory planners have recently shown strong closed-loop performance by iteratively denoising a full-horizon plan, but they remain difficult to certify and can fail catastrophically in rare or out-of-distribution scenarios. To address this challenge, we present PC-Diffuser, a safety augmentation framework that embeds a certifiable, path-consistent barrier-function structure directly into the denoising loop of diffusion planning. The key idea is to make safety an intrinsic part of trajectory generation rather than a post-hoc fix: we enforce forward invariance along the rollout while preserving the diffusion model's intended path geometry. Specifically, PC-Diffuser (i) evaluates collision risk using a capsule-distance barrier function that better reflects vehicle geometry and reduces unnecessary conservativeness, (ii) converts denoised waypoints into dynamically feasible motion under a kinematic bicycle model, and (iii) applies a path-consistent safety filter that eliminates residual constraint violations without geometric distortion, so the corrected plan remains close to the learned distribution. By injecting these safety-consistent corrections at every denoising step and feeding the refined trajectory back into the diffusion process, PC-Diffuser enables iterative, context-aware safeguarding instead of post-hoc repair...
arXiv:2603.13952v3 Announce Type: replace
Abstract: In existing Audio-Visual Speech Enhancement (AVSE) methods, objectives such as Scale-Invariant Signal-to-Noise Ratio (SI-SNR) and Mean Squared Error (MSE) are widely used; however, their correlation with perceived speech quality is often suboptimal and provides limited interpretability for optimization. This work proposes a reinforcement learning-based AVSE framework with a Large Language Model (LLM)-based interpretable reward model. An audio LLM generates natural language descriptions of enhanced speech, which are converted by a sentiment analysis model into a 1-5 rating score serving as the PPO reward for fine-tuning a pretrained AVSE model. Compared with scalar metrics, LLM-generated feedback is semantically rich and explicitly describes speech quality improvements. Experiments on the AVSEC-4 dataset show that the proposed method outperforms a supervised baseline and a DNSMOS-based RL baseline in PESQ, STOI, neural quality metrics, and subjective listening tests.
arXiv:2607.13588v1 Announce Type: new
Abstract: A single-track Gray code is a cyclic Gray code with codewords of length $n$, over an alphabet of size $m$, such that all the $n$ tracks that correspond to the $n$ distinct coordinates of the codewords are cyclic shifts of the first track. Such codes have advantages over the conventional Gray codes in certain quantization and coding applications. Unless $n=2$, there are no such binary codes that contain all the $2^n$ codewords of length $n$. In this paper, we prove that such codes of length $p^t$, $n \geq 2$, with $p^{p^t}$ codewords, over $\F_p$, $p$ prime, exist, for $p=3$ and $p=5$. For larger prime $p$ an appropriate code for $t=2$, implies the existence of such a code for any $t >2$. If the alphabet size $m$ is not a prime there are also indications that such codes exist.
arXiv:2606.03527v2 Announce Type: replace
Abstract: Advertisements reveal information to consumers who decide on further information acquisition and eventual purchase. Anderson and Renault (2006) first modeled this problem using an information-design framework where the advertiser acts as a sender and the consumer as a receiver. Due to search frictions and the consumer's outside option, search for additional information is not always worthwhile for the receiver. Thus, the sender's information design is used to make further search attractive and ultimately induce purchase. Following Lyu (2023), we study the optimal information design for a sender who sells search goods to a consumer with uncertain alternatives. Instead of making relaxations to the sender's problem (as done in Lyu, 2023), we work directly on the joint distribution over realized values and signals. Our contributions are twofold. First, we give a method, based on duality arguments, to verify whether a given information strategy is optimal. We illustrate the value of this verification framework in a competitive extension, where it certifies a non-trivial symmetric equilibrium for two senders with a common convex prior. Second, on the algorithmic front, we develop an FPTAS that finds for the seller a signaling scheme whose utility differs from that of the optimal solution by at most an additive $\epsilon$ error, for any $\epsilon > 0$.
arXiv:2603.12712v2 Announce Type: replace
Abstract: Large language models~(LLMs) have demonstrated remarkable capabilities in code generation, yet their performance remains limited on domain-specific tasks such as Computer-Aided Design~(CAD) code generation, largely due to the scarcity of high-quality training data. In-Context Learning~(ICL) provides a training-free alternative by prompting LLMs with task-specific exemplars, but its effectiveness critically depends on how exemplars are selected. Existing selection strategies mainly rely on similarity or point-wise diversity, often overlooking the compositional nature of CAD design specifications, where a query may involve multiple functional requirements, geometric constraints, and design primitives. As a result, selected exemplars can be individually relevant but collectively redundant, providing insufficient coverage for complex design requirements. In this work, we propose \emph{knowledge sufficiency} as a principled objective for exemplar selection, aiming to select a compact set of exemplars that maximally satisfies the requirements contained in a target design specification. To instantiate this objective, we introduce \emph{Design-Specification Tiling~(DST)}, which estimates knowledge sufficiency through a surrogate tiling ratio by decomposing design specifications into multi-granular components and measuring the proportion of query components covered by selected exemplars. We further show that optimizing this objective can be formulated as a submodular maximization problem, and develop a polynomial-time greedy algorithm tailored to this setting with a $(1-1/e)$-approximation guarantee. Extensive experiments across multiple LLMs demonstrate that DST substantially improves CAD code generation quality and consistently outperforms existing ICL exemplar selection strategies, highlighting the importance of requirement-level knowledge coverage for domain-specific code generation.