arXiv:2512.22867v2 Announce Type: replace Abstract: Socially compliant navigation requires structured reasoning about dynamic pedestrians and physical constraints to ensure safe and interpretable decisions. Vision-language models (VLMs) provide a promising foundation for this task because they can integrate visual observations with language-based social knowledge. However, existing untuned VLMs still struggle to reliably understand fine-grained social norms, making task-specific fine-tuning essential. At the same time, no large-scale egocentric dataset is available this task. To address these challenges, we introduce MUSON, a multimodal dataset for short-horizon social navigation containing 10,110 egocentric samples collected across diverse indoor and outdoor social scenes. MUSON adopts a structured five-step chain-of-thought annotation framework comprising perception, prediction, reasoning, action, and explanation. It explicitly models static physical constraints and employs a standardized six-action decision space. Compared with existing social-navigation datasets, MUSON provides consistent annotations for reasoning, actions, and explanations. We evaluate ten representative small-to-medium VLMs on MUSON. Qwen3-VL-8B achieves the strongest decision-level performance, attaining the highest action accuracy of 0.7765 and Macro-F1 score of 0.7490, as well as the lowest collision rate of 0.0609. These results demonstrate that MUSON is an effective and reusable benchmark for advancing socially compliant navigation. The dataset is publicly available at https://github.com/MUSON-dataset/MUSON/releases/tag/v1.0.
Science Journals
arXiv:2410.17006v4 Announce Type: replace Abstract: Bioacoustic data from Passive Acoustic Monitoring (PAM) generates large datasets where obtaining detailed auditing and labelling is often impractical, resulting in weak annotations (e.g., presence/absence of species over several minutes of recording). In order to effectively capture the complex temporal patterns and key features of long audio segments, we propose a framework comprising dataset standardisation, feature extraction, and classification via Temporal Convolutional Networks (TCN). This approach eliminates the necessity for setting heuristic decision rules or creating time-consuming strong labels. To demonstrate the effectiveness of our approach, we use sperm whale (\textit{Physeter macrocephalus}) click trains in 4-minute recordings as a case study, from a dataset comprising diverse sources and deployment conditions to maximise generalisability. Our TCN classifiers achieve recall rates exceeding 0.83 at a 0.13 false positive rate, comparable to agreement rates between expert annotators. We compare two methods of feature extraction, Variational AutoEncoders (VAEs) and traditional handpicking of features, and found them to yield similar performance results, with the VAE-based classifiers seeing a more stable performance across datasets and recording conditions. These results offer a way forward in leveraging numerous existing annotated bioacoustic datasets to train automatic classification models, effectively overcoming previous limitations associated with weak labels.
arXiv:2502.01044v2 Announce Type: replace Abstract: Drone racing requires high-speed navigation through three-dimensional paths, posing significant challenges in control engineering. Existing control methods lack a feedback control framework that simultaneously addresses nonlinear drone dynamics and multi-agent competitive interactions, such as overtaking or obstructing opponents. To overcome this limitation, this study proposes a game-theoretic control framework, the nonlinear receding-horizon differential game (NRHDG), for competitive drone racing. NRHDG accounts explicitly for adversarial behavior by predicting and countering an opponent's worst-case behavior in real time. It extends standard nonlinear model predictive control (NMPC), which typically assumes a fixed opponent model. First, we develop a novel path-following formulation based on projection-point dynamics, eliminating the need for computationally expensive distance minimization during online control. Second, we propose a potential function that enables each drone to dynamically switch between overtaking and obstructing maneuvers, depending on the race situation. Third, we establish new performance metrics to evaluate NRHDG against NMPC across racing scenarios. Simulation results demonstrate that NRHDG outperforms NMPC in both overtaking and obstructing performance. Specifically, for randomly generated initial conditions and different levels of speed advantage for the rear-start drone, the 95\% confidence intervals for the arc-length-based mean performance differences excluded zero, indicating statistically significant advantages of NRHDG over NMPC in both overtaking and obstructing.
arXiv:2504.13853v2 Announce Type: replace-cross Abstract: Rational design of lipid nanoparticles (LNPs) for tissue-specific delivery critically depends on predicting the composition of the protein corona that forms on the lipid surface after intravenous administration. However, conventional characterization of the protein corona relies on costly and time-consuming mass spectrometry experiments, which require physically prepared liposome samples and therefore cannot serve as a pre-synthesis screening strategy for large candidate lipid spaces. The adsorption of plasma proteins onto liposomal surfaces is shaped by lipid chemical structures, protein properties and the biological environment, making this process difficult to simulate directly. In this work, we propose that scoring lipid-plasma protein pairs and ranking the resulting scores can provide a practical signal for revealing the relative composition of the liposomal surface protein corona.Here we introduce GenShin, a geometry-enhanced pose-free graph neural network designed to score lipid-plasma protein pairs. GenShin is pretrained on compound-protein affinity data to initialize a generalizable scoring function and is then fine-tuned on a rank fine-tuning dataset constructed from liposomal protein-corona abundance measurements to adapt the model to lipid-plasma protein pair scoring. Before fine-tuning, GenShin achieves competitive pose-free affinity prediction on the PDBbind v2016 benchmark compared with representative pose-dependent models. CASF-2016 perturbation experiments using the pretrained GenShin model further show that pose-dependent inference substantially degrades when intermolecular poses are unreliable, whereas GenShin remains stable without requiring such poses. This supports the practical advantage of GenShin for large-scale lipid-protein scoring.
arXiv:2404.17497v2 Announce Type: replace Abstract: We study how bug bounty programs (BBPs) shape software vendors' security and release choices. We develop a game-theoretic model in which a vendor chooses release timing and severity-contingent bounties, anticipating effort by ethical and malicious hackers in a winner-take-all discovery race. The model highlights two linked mechanisms: an incentive channel that shifts first discovery of severe vulnerabilities away from malicious exploitation and toward ethical reporting, and a governance channel in which coordinated disclosure changes how vulnerability information is managed during remediation. We derive closed-form optimal bounties and characterize a feasibility region sustaining positive bounties and interior success probabilities. Within it, a BBP strictly increases the vendor's expected profit by reallocating first-discovery probability on severe vulnerabilities from malicious to ethical hackers and by converting part of severe-loss exposure into bounded, pay-for-results expenditures. For private programs, we solve for the optimal invited set of ethical hackers and show it is strictly smaller than the expected number of malicious attackers. Higher bounties raise ethical hackers' effort and first-discovery probabilities but also increase program cost, and interact with reputational (non-monetary) incentives. Finally, BBP adoption conditionally reduces the marginal value of additional pre-release delay, implying earlier release relative to the no-BBP benchmark. Managerially, BBPs should be viewed as a post-release governance layer complementing strong internal assurance rather than a substitute for it. Policymakers can support responsible use by encouraging timely remediation, transparent post-patch disclosure, and reporting standards that reduce information asymmetry and triage frictions. (Abstract edited to fit the arXiv length limit.)
arXiv:2506.13385v2 Announce Type: replace Abstract: Human mobility shapes access to resources, opportunities, and services, making movement data a powerful lens for studying spatial and social inequality. Yet despite the growing availability of official open mobility datasets, their research potential is rarely realized because the technical overhead of retrieving, harmonizing, and processing them often crowds out substantive analysis. To address this, we introduce pySpainMobility, a Python package that automates the retrieval and harmonization of Spain's open mobility data across spatial resolutions and demographic strata, streamlining national-scale, reproducible analysis. Using the package, we study income-stratified mobility inequality across Spain's inter-province network, drawing on district-level origin-destination flows for four representative weeks spanning the seasons of 2023. We construct income-specific mobility layers and show that socioeconomic stratification is deeply embedded in the structure of the national mobility system: low-income mobility is disproportionately concentrated in a narrow set of destinations and shorter in spatial reach, while high-income groups access a broader and more distant hierarchy of destinations. Low- and high-income layers consistently follow weakly aligned destination hierarchies across seasons, indicating that income groups navigate distinct mobility geographies rather than a shared one at different volumes. We further show that destination provinces themselves differ systematically in the income composition of the travelers they receive, with several provinces attracting arrivals disproportionately skewed toward one income group relative to the national seasonal baseline. These results demonstrate how official open mobility data, combined with accessible tooling, can be operationalized to reveal spatial inequality as a structural property of national mobility networks.
arXiv:2606.29580v3 Announce Type: replace Abstract: Maternal and newborn mortality remain among the highest in sub-Saharan Africa, where midwifery care is often delivered by nurses who lack midwifery training to international standards, and consulting authoritative guidance at the point of care is hard: the guidelines are long and connectivity is intermittent. We present MAM-AI, a medical question-answering assistant for nurse-midwives in Zanzibar that runs entirely on a commodity Android device: a question is embedded (EmbeddingGemma, 300M) and matched against a curated corpus of 87 guideline documents (63,650 passages), then answered with citations by a 4B int4 generator (Gemma 4 E4B), fully offline, with no query leaving the device. We evaluate the exact deployed configuration with a layered methodology -- retriever, generator under oracle context, end-to-end, and latency -- scored by LLM judges validated against physician rubrics. The evaluation relocates the hard problem. On-device retrieval is essentially solved: the 300M embedder ranks third of seven retrievers and rivals cloud systems, so the passages the system needs are usually found. The small generator is what remains in doubt: adding retrieved context does not improve its answers, and at 4B it cannot be both helpful and safe at once -- of two same-size candidates, the more helpful one commits genuine dangerous errors, so we deploy the other, which is about twice as faithful to its sources (as faithful as a frontier model), and recover its helpfulness with a redesigned prompt that cuts deflection from 33% to 3%. Corpus quality is decisive for the same reason: where the corpus holds the right passage the answer is specific and actionable, and where it does not it goes vague. MAM-AI is a thoroughly evaluated, open-source research prototype, not a fielded product; the system, knowledge base, benchmarks, and evaluation harness are released.
arXiv:2606.07322v2 Announce Type: replace-cross Abstract: The traveling salesman problem (TSP) is a classic NP-hard problem. Held--Karp dynamic programming~\cite{held1962dynamic, bellman1962dynamic} solves it exactly in $O(n^2 2^n)$ time, a barrier that has stood for over six decades. Whether quantum computing can surpass $O^*(2^n)$ is a central open question. The authors of~\cite{ambainis2019quantum}\ (SODA~2019) claimed a query complexity of $O^*(1.727^n)$, but we identify a structural counting error: when corrected, their scheme requires $\Omega^*(2^n)$ queries and offers no advantage over classical Held--Karp. We design a quantum divide-and-conquer framework: partition $n$-vertex set into $k$ subsets, classically precompute shortest paths within each, then search over all $k$-partitions via quantum minimum finding. We prove $k=3$ achieves $O^*(1.890^n)$, and $k=4$ attains the global optimum $O^*(1.866^n)$, the first quantum algorithm to surpass $O^*(2^n)$ for general TSP. To convert the query bound into a time-complexity advantage, we overcome the oracle-construction bottleneck by preparing a set partition state, requiring $O(n^2)$ gates and $O(n)$ depth. Leveraging structured state preparation, we achieve a total time complexity of $O^*(1.866^n)$. Qiskit simulations on $n=6,7$ achieve $98.9\%$ and $100\%$ accuracy.
Weighted Phase Volume Method in Stability Analysis: Integral Criteria and Ellipsoidal Reachable Sets
arXiv:2607.05033v1 Announce Type: new Abstract: A method for analysing the stability of dynamical systems is proposed, based on the introduction of a weighted phase volume and time rescaling by a positive function. The advantage of the method is the ability to set the contraction properties of the phase volume by choosing the weighting function and the scaling factor, while preserving the topology of the phase portrait. Integral dissipativity conditions are derived, leading to new definitions of integral stability, asymptotic stability, and exponential stability. For quadratic weighting functions, covering and inner ellipsoids are constructed, providing geometric estimates of reachable sets. The connection between the proposed approach and classical Lyapunov stability is established. The efficiency of the method is demonstrated through numerical examples.
arXiv:2605.09872v3 Announce Type: replace-cross Abstract: It is known that there exist multi-prover interactive protocols ($\mathsf{MIP}$ protocols) for the complexity class $\mathsf{NEXP}$, succinct $\mathsf{MIP}$ protocols for $\mathsf{NP}$ and multi-prover interactive protocols with shared entanglement ($\mathsf{MIP}^\ast$ protocols) for $\mathsf{RE}$. This extraordinary power of multi-prover interactive proof systems comes from the assumption that provers do not communicate with each other during the protocols. If they are allowed to communicate freely, the setting is the same as in the single-prover case, and the computational power of the system becomes significantly weaker. In this paper, we investigate for the first time the setting where communication (i.e., leakage of information) between provers is allowed but bounded. We introduce two techniques to approach this question and show that multi-prover interactive proof systems are robust against some amount of leakage. Our first technique is based on parallel repetition theorems. We apply it to show that for any polynomial $p$, we can construct two-prover one-round $\mathsf{MIP}$ and $\mathsf{MIP}^\ast$ protocols for $\mathsf{NEXP}$ and $\mathsf{RE}$, respectively, that are robust against $p(n)$ bits of leakage. We further derive our second technique to convert any low-soundness PCP construction to a two-prover one-round $\mathsf{MIP}$ protocol for $\mathsf{NP}$ robust against leakage. We also discuss the relation between robustness against leakage in multi-prover interactive proof systems and the Sliding Scale Conjecture in the PCP literature.
arXiv:2506.07406v3 Announce Type: replace Abstract: Understanding the internal representations of large language models (LLMs) is a central challenge in interpretability research. Existing feature interpretability methods often rely on strong structural assumptions--such as linearity or sparsity--that may not hold in practice. In this work, we introduce InverseScope, an assumption-light and scalable framework for interpreting neural activations via input inversion. Given a target activation, InverseScope characterizes its encoded information by generating natural-language inputs that produce nearby activations, grounding abstract internal states in concrete language. To overcome the prohibitive cost of sampling in high-dimensional activation spaces, we propose a novel control-layer conditioning architecture that substantially improves sample efficiency compared to prior token-prepending approaches. We demonstrate that InverseScope reveals rich geometric structure in LLM representation spaces, including sentence-level linear analogies. The framework scales to state-of-the-art open-source models of up to 14B parameters and generalizes to out-of-distribution inputs, enabling systematic analysis of activation neighborhoods.
arXiv:2506.16898v2 Announce Type: replace Abstract: Diffusion-based text-to-image models are increasingly used for urban analysis and scenario generation, but their geographic knowledge and representational biases remain poorly understood. We evaluate FLUX 1-schnell and Stable Diffusion 3.5-Large in the United States by generating 150 street-view images for each state, each state capital, and a generic "USA" prompt. Images are embedded with DINO-v2 ViT-S/14 and compared with Fr\'echet Inception Distance (FID). Pairwise FID clustering shows that geographically proximate states and capitals often group together, indicating implicit geographic structure. However, the generic ``USA'' prompt collapses this diversity into a metropolitan stereotype: frontier, desert, tropical, rural, and small-city environments are underrepresented or distant in FID space. These results show that diffusion models can encode fine-grained geography while still reproducing narrow national-scale visual stereotypes.
arXiv:2508.06577v3 Announce Type: replace Abstract: Participatory budgeting is a democratic innovation that allows citizens to propose and vote on public investment projects. To help organizers manage large volumes of submissions, we design and test privacy-preserving methods for algorithmic shortlisting. These algorithms predict which projects are likely to be funded using only project features and anonymous historical voting data. We demonstrate the limitations of a naive approach that uses a large language model to rank projects based on past success and propose a vote-based pipeline that enables state-of-the-art LLMs to perform on par with classical machine learning. Our findings indicate that user preferences in participatory budgeting are stable enough to allow algorithmic shortlisting to approximate an initial selection of projects effectively.
arXiv:2607.05271v1 Announce Type: new Abstract: Physics-informed neural networks (PINNs) encounter ill-posed optimization, loss competition, and parameter compensation in partial differential equation (PDE) inverse problems. Transfer learning can reuse representations from source tasks, but direct fine-tuning may introduce negative transfer when dominant physical mechanisms, governing parameters, or observation noise differ between source and target domains: the model achieves low field error yet recovers incorrect target physical parameters. To mitigate, we propose Target-Guided Selective Reweighting PINN (TGSR-PINN), a target-evidence-driven representation correction method for PINN inverse transfer learning. TGSR-PINN transfers only the weights and biases from the source PINN, while target physical parameters are independently initialized; after a short target-adaptation phase, the method computes neuron target scores using first-order Taylor sensitivity and pre-activation variance on fixed scoring batches, and converts evidence associated with low-scoring neurons into continuous weak-adaptation signals via a Gaussian mixture model (GMM) with rank fallback. TGSR-PINN then applies selective soft decay to input weight rows and biases of low-scoring neurons instead of hard pruning or random resetting. In experiments, TGSR-PINN improves target parameter recovery while maintaining comparable field accuracy in the high-P\'{e}clet 2D advection-diffusion task and in the Allen--Cahn to Burgers cross-PDE-family transfer task; a 5%-noise reaction--diffusion case provides supplementary evidence under milder source-target mismatch. Ablation studies suggest that neuron target scoring, weak-adaptation signal estimation, layer protection, and selective soft decay jointly contribute to the benefits.
arXiv:2409.01102v3 Announce Type: replace Abstract: SQL/PGQ and GQL are very recent international standards for querying property graphs: SQL/PGQ specifies how to query relational representations of property graphs in SQL, while GQL is a standalone language for graph databases. The rapid industrial development of these standards left the academic community trailing in its wake. While digests of the languages have appeared, we do not yet have concise foundational models like relational algebra and calculus for relational databases that enable the formal study of languages, including their expressiveness and limitations. At the same time, work on the next versions of the standards has already begun, to address the perceived limitations of their first versions. Motivated by this, we initiate a formal study of SQL/PGQ and GQL, concentrating on their concise formal model and expressiveness. For the former, we define simple core languages -- Core GQL and Core PGQ -- that capture the essence of the new standards, are amenable to theoretical analysis, and fully clarify the difference between PGQ's bottom up evaluation versus GQL's linear, or pipelined approach. Equipped with these models, we both confirm the necessity to extend the language to fill in the expressiveness gaps and identify the source of these deficiencies. We complement our theoretical analysis with an experimental study, demonstrating that existing workarounds in full GQL and PGQ are impractical which further underscores the necessity to correct deficiencies in the language design.
arXiv:2606.28895v2 Announce Type: replace-cross Abstract: We investigate linear lumping for parameter-dependent mass action reaction networks, distinguishing between generic and critical parameter regimes. For generic parameters -- those ranging in some non-empty open subset of parameter space -- we prove that exact linear lumping yields only "obvious" reductions: elimination of non-reactant species or projections along stoichiometric first integrals. This characterization extends to reaction networks with product-form kinetics, including Michaelis-Menten and Hill-type rate laws. For mass action systems we proceed to develop an algorithmic approach to identify critical parameter sets -- algebraic subvarieties in parameter space where non-trivial lumpings become available. This procedure reduces the determination of lumping maps to a system of finitely many polynomial equations. It also applies to constrained lumping scenarios (which are frequently motivated by chemical considerations). We then review and extend results about proper lumpings. Finally, we discuss lumpings of a self-replicator system, and of a two-pathway enzyme mechanism, to document the viability of our methods in relevant scenarios. Our results clarify the relationship between structural (parameter-independent) and fine-tuned (parameter-dependent) reductions, with implications for approximate lumping when system parameters lie near critical values
arXiv:2603.06577v2 Announce Type: replace Abstract: While recent multimodal large language models (MLLMs) have made impressive strides, they predominantly employ a conventional autoregressive architecture as their backbone, leaving significant room to explore effective and efficient alternatives in architectural design. Concurrently, recent studies have successfully applied discrete diffusion models to various domains, such as visual understanding and image generation, revealing their considerable potential as a promising backbone for multimodal systems. Drawing inspiration from these pioneering studies, we introduce Omni-Diffusion, the first any-to-any multimodal language model built entirely on mask-based discrete diffusion models, which unifies understanding and generation across text, speech, and images. Omni-Diffusion employs a unified mask-based discrete diffusion model to directly capture the joint distribution over discrete multimodal tokens. This approach supports not only bimodal tasks but also more complex scenarios involving multiple modalities. On a diverse set of benchmarks, our method outperforms or performs on par with existing multimodal systems that process two or more modalities, highlighting the significant promise of diffusion models in powering the next generation of multimodal foundation models. Project webpage: https://omni-diffusion.github.io.
arXiv:2603.06921v2 Announce Type: replace Abstract: Safe navigation of autonomous robots remains one of the core challenges in the field, especially in dynamic and uncertain environments. One prevalent approach is safety filtering based on control barrier functions (CBFs), which are easy to deploy but difficult to design. Motivated by the shortcomings of existing learning- and model-based methods, we propose a simple yet effective neural CBF design method for safe robot navigation in dynamic environments. We employ the idea of a composite CBF, where multiple neural CBFs are combined into a single CBF. Individual CBFs are trained using data generated offline via the Hamilton-Jacobi reachability framework to approximate the optimal safe set for single moving obstacles. Additionally, we use a residual neural architecture, ensuring that the estimated safe set does not intersect with the corresponding failure set. The method is extensively evaluated in simulation experiments for a ground robot and a quadrotor, comparing it against several baseline methods. The proposed method improves success rates by up to 18\% over the strongest baseline, while maintaining comparable or lower path lengths and motion times. The method is also demonstrated in hardware experiments for both types of robots.
arXiv:2607.05346v1 Announce Type: new Abstract: We propose OptiAgent, a multi-agent framework that, given a natural language description of an Operations Research problem, is able to output a solver-ready mathematical formulation as well as executable code. Our architecture prioritizes the mathematical modeling step, where dedicated agents extract structures, such as decision variables and constraints, enabling iterative self-correction. We introduce a novel multi-loop validation architecture with four specialized feedback mechanisms, each targeting a distinct failure mode such as misinterpretation, structural defects, mathematical inconsistencies, validation failures, and code errors. Alongside accuracy, our modular design improves the process of solving optimization problems by improving transparency, as each agent exposes its reasoning and feedback, making the full modeling process auditable. Our framework achieves state-of-the-art performance on 3 out of 4 benchmarks across LP, MILP, and Nonlinear Programming tasks, while remaining highly competitive on the remaining dataset.
arXiv:2607.05325v1 Announce Type: new Abstract: Cerebral microbleeds (CMBs) are MRI markers of small vessel disease and the microbleed component of amyloid related imaging abnormalities (ARIA-H), but their small size, sparsity, and similarity to vessels, calcification-like foci, and artefacts make automated detection difficult. We propose CenSynCMB, a centre-guided and mimic-aware framework combining a 3D Attention U-Net, auxiliary centre-map supervision, false-negative-driven reweighting, and fold-wise physics-guided synthesis of positive CMBs and labelled hard negatives. Synthetic data expose the detector to compact lesions and common mimics without validation or test leakage. On VALDO Task 2, CenSynCMB achieved the best local-comparison lesion-level F1 (74.3%, p = 0.020); on external AIBL SWI, it achieved the highest local-comparison recall (88.5%, p = 0.0058) and F1 (65.0%, p = 0.0016). Together, these results support scalable CMB candidate extraction in large, unlabelled MRI cohorts, while highlighting cohort-specific calibration as the next step toward reliable burden estimation.
arXiv:2607.05365v1 Announce Type: new Abstract: Streaming speech-to-speech language models aim to answer spoken queries directly with synthetic speech. However, standard speech and text benchmarks do not capture whether these systems behave naturally in conversations, where timing, turn-taking, prosody, interpersonal stance, language and dialect consistency, and relationship-aware appropriateness jointly shape perceived quality. We introduce SPEARBench, a benchmark for evaluating naturalness in speech-to-speech language models from question-answer interactions. SPEARBench constructs controlled dialogue prompts from the Seamless Interaction corpus, runs inference across multiple models, and evaluates generated answers using a multidimensional protocol that covers response latency, interruptions, speech quality, ASR robustness, language and dialect consistency, emotional naturalness, interpersonal stance, and explainable distributional baselines. The benchmark includes original human answers as a reference condition and reports results for several contemporary models. Results show that current models can achieve high signal-level quality and low ASR error while still differing from human conversational behavior in latency, overlap, dialect preservation, emotional adaptation, and interpersonal stance dynamics.
arXiv:2607.05387v1 Announce Type: new Abstract: We propose a protocol for cross-chain atomic transactions (CATs), enabling composable atomic execution across different blockchains. The protocol addresses the key interoperability challenge of providing atomicity guarantees in the presence of asynchronous communication and Byzantine actors. It preserves chain autonomy by allowing each blockchain to maintain its own execution model while participating in coordinated cross-chain operations. The design introduces a shared coordination layer involving sequencers, transaction processors, a coordinator, and a confirmation layer which together ensure that either all parts of a CAT succeed or none do. To prevent unnecessary blocking, we separate transaction execution into accepted and postponed sets, with the coordination layer resolving the outcomes of CATs within a few rounds. We further introduce timeouts and dependency-depth bounds for liveness and mitigation of cascading delays. Our formal analysis establishes strong safety and liveness guarantees and demonstrates that the protocol achieves minimal blocking for independent transactions while ensuring bounded blocking time for dependent transactions. Experimental evaluation shows high CAT success when cross-chain transactions are a modest share of traffic, and characterizes the CAT-lifetime trade-off between success and dependent-transaction latency. This protocol enables fast, secure, and deterministic atomic cross-chain execution while preserving chain autonomy, providing a foundation for scalable blockchain interoperability solutions.
arXiv:2607.02620v1 Announce Type: cross Abstract: Simulating molecules is a major application of quantum computing, with the potential to overcome exponential scaling constraints of classical computation. Researchers use different methods in order to evaluate the readiness of NISQ computers in order to test current simulation capabilities. We present an integrated repository with reproducible benchmarks of over 10 different ansatzes from published papers and two different truncation methods, applicable to any set of mapped hamiltonians, providing a single pipeline for comparing performance along multiple axes, including variance and computational time, among others. We apply them to simulate different amino acids, using hamiltonians taken from the QMProt Dataset. We then ran four separate experiments. First, we quantified noise resilience by optimizing the same hardware-efficient ansatzes under identical initialization while sweeping PennyLane noise channels and strengths, and measuring parameter drift, cosine similarity of optimal parameters, and energies evaluated on noiseless versus noisy backends. We then studied barren-plateau-related trainability via gradient-variance diagnostics and optimization trajectories across initialization strategies and ansatzes depth on small systems. We then compared adaptive versus fixed ansatzes at matched parameter budgets, reporting outer-loop iterations, wall time, and especially total cost-function evaluations to fairly contrast greedy adaptive growth with layered hardware-efficient circuits. Lastly, we mapped accuracy versus expressive capacity by sweeping the number of retained adaptive operators and recording ground-state energy error relative to classical references.
arXiv:2605.02504v2 Announce Type: replace Abstract: Most hallucination evaluations focus on English, leaving it unclear whether findings transfer to lower-resource languages. We investigate faithfulness hallucinations, defined as model-generated content that is fluent and plausible but diverges from the provided input or is internally inconsistent. Leveraging the multilingual MultiWikiQA dataset, we utilize the LettuceDetect framework to create synthetic hallucination datasets for 306 languages, from which we train token-level hallucination classifiers for 30 European languages. In this work, we present evaluations of model hallucinations on a selection of languages: English, Danish, German, and Icelandic. Using these classifiers, we evaluate the hallucination rates for Qwen3-0.6B, Qwen3-14B, Gemma-3-12B-IT, cogito-v1-preview-qwen-32B, and cogito-v1-preview-llama-70B. Our classifiers reveal notably higher hallucination rates for Qwen3-0.6B (up to 60\% of answers containing at least one hallucination, peaking in Icelandic) and generally lower rates for larger models, with cogito-v1-preview-qwen-32B and cogito-v1-preview-llama-70B performing best on most languages. Hallucination rates are consistently higher for lower-resource languages, particularly Icelandic.
arXiv:2605.04998v3 Announce Type: replace Abstract: This revision updates a pop-to-jazz chord-generation rehearsal study. Best-epoch metrics still show that modest pop rehearsal preserves pop accuracy while improving jazz prediction, but v2 corrects released-checkpoint selection: the released F1 equals Phase 0, F2 had a transcription error, and ft-pop80-v2 restores a hash-distinct jazz-adapted F1 across 3 seeds.