arXiv:2606.08367v1 Announce Type: new Abstract: Most evaluations of LLM agents look like exams: a discrete task, a clean environment, a score in minutes or hours. We argue that this approach is mismatched with the deployment conditions of autonomous systems, where the relevant timescale can be weeks to months, and where the dynamics that matter most, such as behavioral drift, governance in diverse environmental contexts, and cross-influence between agents from different model families, only emerge over time. We introduce Emergence World, a continuously running multi-agent simulation platform designed to make those dynamics measurable. The platform hosts populations of LLM-driven agents in a shared spatial world grounded in live external data (e.g. real-time weather, news APIs, internet access), equips each agent with 120+ specialized tools and three persistent memory systems, and lets them govern themselves through democratic mechanisms with consequential outcomes. The platform is model-agnostic at the reasoning layer and supports heterogeneous populations in which agents from different vendors share the same world. To illustrate the kinds of questions the platform makes tractable, we present a 15-day cross-vendor study with five parallel worlds powered by Claude Sonnet 4.6, Grok 4.1 Fast, Gemini 3 Flash, GPT-5-mini, and a mixed population. Identical roles and starting conditions produced radically different outcomes, ranging from stable deliberative governance to total population collapse. We release the prompts, log data and configurations to support further research on long-horizon multi-agent autonomy.
Science Journals
arXiv:2606.09674v1 Announce Type: new Abstract: We present Trellis: an autoformalization system that leverages LLM agents in a deterministically constrained workflow to enforce incremental progress in Lean autoformalization tasks through iterative refinement of natural language proofs. Our approach is motivated by the common mathematician's notion of what it means to have a rigorous proof in the first place: namely, that it would be routine to elaborate any part of the proof in further detail. The result is a system which aims to achieve reliable autoformalization on a modest budget and with generalist agents, with specialization to autoformalization coming not from any task-specific agent training but instead from a meaning-of-rigor inspired workflow enforced by process semantics. We link to an end-to-end Lean formalization of a recent Ramsey theory breakthrough produced by the process.
arXiv:2606.09688v1 Announce Type: new Abstract: We develop a fast space-time numerical scheme for approximating solutions to the linear wave equation. The approach is based on the sparse-grid combination technique applied to a coercive space-time discretization. Designed for tensor-product space-time discretizations, the method enables efficient parallelization of the resulting solver. We provide a rigorous theoretical analysis establishing convergence rates and computational complexity estimates. Numerical experiments validate the theoretical estimates and demonstrate the efficiency of the proposed method.
arXiv:2606.08794v1 Announce Type: cross Abstract: Adaptive variational quantum algorithms like ADAPT-VQE construct tailored ans\"atze by iteratively selecting operators from a pool using gradient-based criteria. While this avoids oversized parameter spaces, repeatedly scanning the full pool incurs a classical cost that scales linearly with pool size-a major bottleneck for systems with long-range interactions or large operator sets. Here, we reformulate adaptive operator selection as a graph-based decision problem and introduce a graph neural network (GNN) policy that predicts the next entangling operator directly from the interaction graph and state-dependent observables. Training data are generated from exact simulations of disordered long-range spin chains, using gradient magnitudes as supervision signals. The learned policy accurately reproduces the dominant structure of the greedy gradient-based selection rule, significantly outperforming heuristics based solely on interaction strength. Integrated into a variational quantum eigensolver (VQE) workflow, this GNN-VQE approach achieves energy errors close to standard ADAPT-VQE while drastically reducing full-pool gradient evaluations. To test transferability beyond spin models, we evaluate the policy on small active-space molecular benchmarks (LiH and BeH_$2$). We find the GNN is highly effective as a shortlist generator: exact rescoring over just a few GNN-proposed candidates recovers near-oracle rollout behavior while searching only a small fraction of the pool. These results demonstrate that adaptive circuit construction contains learnable structure that can be exploited to accelerate variational quantum algorithms.
arXiv:2601.01665v2 Announce Type: replace Abstract: Deep reinforcement learning (DRL) has shown great promise in addressing multi-objective combinatorial optimization problems (MOCOPs). Nevertheless, the robustness of these learning-based solvers has remained insufficiently explored, especially across diverse and complex problem distributions. In this paper, we propose a unified robustness-oriented framework for preference-conditioned DRL solvers for MOCOPs. Within this framework, we develop a preference-based adversarial attack to generate hard instances that expose solver weaknesses, and quantify the attack impact by the resulting degradation on Pareto-front quality. We further introduce a defense strategy that integrates hardness-aware preference selection into adversarial training to reduce overfitting to restricted preference regions and improve out-of-distribution performance. The experimental results on multi-objective traveling salesman problem (MOTSP), multi-objective capacitated vehicle routing problem (MOCVRP), and multi-objective knapsack problem (MOKP) verify that our attack method successfully learns hard instances for different solvers. Furthermore, our defense method significantly strengthens the robustness and generalizability of neural solvers, delivering superior performance on hard or out-of-distribution instances.
arXiv:2601.02085v3 Announce Type: replace Abstract: Strawberry-harvesting robots faced challenges such as poor visual perception, gripper misalignment, empty grasp/misgrasp, and slippage, which reduced harvesting stability and efficiency.To overcome these issues, this paper proposes a visual fault diagnosis and self-recovery framework. An end-to-end SRR-Net achieved unified perception and fault diagnosis through joint detection, segmentation, and ripeness regression of the fruit and gripper. Leveraging this integrated perception, a relative error compensation method driven by simultaneous target-gripper detection was designed to correct positional misalignments exceeding the tolerance threshold. A micro-optical camera integrated within the end-effector delivered real-time visual feedback. Based on the micro-optical camera, a MobileNet V3-Small classifier was utilized for grasp adjustment during the deflating stage, enabling the early abort of the harvesting cycle in cases of empty grasp/misgrasps. Furthermore, a time-series LSTM classifier was applied during the snap-off stage to predict strawberry slippage. Based on these predictions, the system executed re-inflation and a secondary snap-off attempt for slipping strawberries, or aborted the cycle for slipped strawberries. Experiments demonstrated that the mean absolute errors between the end-effector and the picking point were reduced to 3.12 mm and 4.06 mm from 11.50 mm and 5.25 mm along the x- and y-axes, respectively, at the cost of a time increment of 0.64 $pm$ 0.24 s. The grasp adjustment module reduced the grasping phase by approximately 0.5 s and avoided empty-placement for failure cases. The strawberry slip prediction module handled slipped cases with an 88.89% success rate, saving approximately 4.00 s per harvesting cycle for failure cases. Also, it achieved an 81.25% recovery rate for slipping strawberries, requiring additional 0.63 s for re-grasping.
arXiv:2605.26323v3 Announce Type: replace Abstract: Federated Learning (FL) is an emerging distributed machine learning (ML) technique that enables in-situ model training and inference on decentralized edge devices. We propose Totoro$^+$, a novel scalable FL system that enables massive FL applications to run simultaneously on edge networks. The key insight is to explore a distributed hash table (DHT)-based peer-to-peer (P2P) model to re-architect the centralized FL system design into a fully decentralized one. In contrast to previous studies where many FL applications shared one centralized parameter server, Totoro$^+$ assigns a dedicated parameter server to each application. Any edge node can act as any application's coordinator, aggregator, client selector, worker (participant device), or any combination of the above, thereby radically improving scalability and adaptivity. Totoro$^+$ introduces three innovations to realize its design: a locality-aware P2P multi-ring structure, a publish/subscribe-based forest abstraction, and a game-theoretic path planning model with a guarantee of an $\epsilon$-approximate Nash equilibrium. Real-world experiments on 500 Amazon EC2 servers show that Totoro$^+$ scales gracefully with the number of FL applications and $N$ edge nodes speeds up the total training time by $1.2\times-14.0\times$, achieves $\mathcal{O}(\log N)$ hops for model dissemination and gradient aggregation with millions of nodes, and efficiently adapts to the practical edge networks and churns.
arXiv:2502.09194v1 Announce Type: cross Abstract: Generative Artificial Intelligence (AI) techniques have become integral part in advancing next generation wireless communication systems by enabling sophisticated data modeling and feature extraction for enhanced network performance. In the realm of open radio access networks (O-RAN), characterized by their disaggregated architecture and heterogeneous components from multiple vendors, the deployment of generative models offers significant advantages for network management such as traffic analysis, traffic forecasting and anomaly detection. However, the complex and dynamic nature of O-RAN introduces challenges that necessitate not only accurate detection mechanisms but also reduced complexity, scalability, and most importantly interpretability to facilitate effective network management. In this study, we introduce the XAInomaly framework, an explainable and interpretable Semi-supervised (SS) Deep Contractive Autoencoder (DeepCAE) design for anomaly detection in O-RAN. Our approach leverages the generative modeling capabilities of our SS-DeepCAE model to learn compressed, robust representations of normal network behavior, which captures essential features, enabling the identification of deviations indicative of anomalies. To address the black-box nature of deep learning models, we propose reactive Explainable AI (XAI) technique called fastshap-C.
arXiv:2606.09626v1 Announce Type: cross Abstract: We present analytical model for peristaltic transport within compressible, ideal magnetohydrodynamics (MHD). By employing small-amplitude perturbation expansion, under thin-tube long-wavelength approximation with a uniform axial background magnetic field, we study non-linear coupling between thermodynamic pressure variations and Maxwell's magnetic tension stresses. The resulting net time-averaged volumetric flow rate $\langle Q \rangle$ is calculated. When applied to solar chromospheric spicules under equipartition constraints ($\beta \sim 1$), where sound speed matches the Alfv{\'e}n speed, we find $\langle Q \rangle = 4\epsilon^2/(M^2-1)$. Because the denominator remains positive across all operational supersonic Mach numbers ($M \approx 2\text{--}10$), upward-propagating mechanical disturbances drive a highly directional, collimated upward flow which we interpret as a spicule. Estimates show that for observationally realistic magnetosonic waves with amplitudes of $\approx 10\%$, the peristaltic mechanism generates a localized mass flux $\approx 100$ times that of solar wind. We propose an explicit observational signature of this mechanism, wherein the launch of individual spicular jets is directly preceded by magnetosonic wave trains detectable as localized intensity modulations. Beyond solar chromospheric application, the model may be applicable to traveling magnetic field pinches in laboratory plasma devices and astrophysical mass-loading processes in stellar winds and inner regions of magnetized accretion disks.
arXiv:2606.08430v1 Announce Type: new Abstract: Digital Compute-in-Memory (DCiM) reduces data movement and has become a promising solution for energy-efficient edge AI. However, most existing DCiM frameworks still primarily target integer or fixed-point arithmetic, and provide limited support for compiler-integrated and accuracy-configurable floating-point computation. Directly integrating conventional IEEE 754 floating-point units into dense SRAM-based DCiM arrays, however, incurs high area and power overhead. To address this challenge, this work presents an accuracy-configurable floating-point multiplier integrated into the OpenACM framework for SRAM-based DCiM. An exact IEEE~754-compliant multiplier is first implemented as a baseline, and a mantissa-segmentation-based approximate multiplier is then proposed to reduce hardware cost while preserving numerical fidelity. Post-layout results show up to 69% logic area reduction and 72% power savings over exact floating-point designs without delay overhead. Evaluations on image processing tasks and ResNet-18 inference further demonstrate negligible accuracy degradation. These results indicate that compiler-integrated approximate floating-point multiplication is a practical approach for enabling efficient and configurable floating-point support in SRAM-based DCiM systems. The Floating-Point Multiplier is available on https://github.com/ShenShan123/OpenACM
arXiv:2606.08311v1 Announce Type: new Abstract: Electronic health record (EHR) notes are dense medical documents containing large amounts of information, often filled with complex medical jargon. Highlighting all details in EHRs helps reduce the likelihood of missing crucial information by drawing attention to key content. This study proposes the design of a Cardiology Interface Terminology (CIT) to accurately highlight all details in EHR notes of cardiology patients. We introduce an innovative Machine Learning (ML) technique for the design of CIT. The ML technique requires training data. Manual preparation of such training data is time-consuming and expensive. The process of the CIT design includes three phases. In the first two phases, we innovatively derive a training data CIT to be used by the third phase, ML technique. We start by designing an initial CIT, composed of several components: the cardiology-related sub-hierarchies of SNOMED, other SNOMED concepts mined from EHRs of build set, and necessary components of terms e.g., medical abbreviations and medications. Utilizing an iterative process, fine-grained phrases containing initial CIT concepts are extracted from build set as CIT concept candidates. The candidate concepts are semi-automatically reviewed before being added to CIT, yielding the training data CIT, TCIT. In the third phase, a ML model is trained with TCIT to identify candidates fitting to be concepts in the CIT. This model is used to extract further concepts from build set, yielding the final CIT. The final CIT is then used to highlight the test set and evaluate the extent to which it captures details in an unseen EHR dataset. For this purpose, four evaluation metrics, coverage, breadth, completeness, and conciseness are used. The highlighted test set has a coverage of 74.21%, with a breadth of 1.68. For 20 random notes in test set, the average completeness is 98.2% and average conciseness is 84.2%.
arXiv:2606.08434v1 Announce Type: new Abstract: The shift to renewable-dominated power systems has produced low-inertia grids, undermining system stability. In this context, grid-forming inverters (GFMs) have emerged as a promising solution. However, GFMs challenge conventional analysis techniques, especially those relying on small-signal or root-mean-square (RMS) models. Such models rely on linearization and sinusoidal steady-state assumptions, which fail in large-signal cases. Stability of GFM-based systems therefore becomes operating-point dependent, and a feasible operating point may not even exist. While large-signal analyses are available, decentralized certification of operating-point convergence with explicit transient guarantees, such as rate and overshoot, remains rare. This paper proposes an algebraic, decentralized contraction-based framework. The proposed contraction stability analysis certifies system stability and convergence to desired operating points. The method works in the time domain and captures nonlinear, large-signal behavior of synchronization and power-sharing mechanisms. Moreover, the contraction rate provides an explicit bound on transient time: trajectories converge exponentially to the new operating point at a controlled rate, yielding computable contraction regions that certify stability and large-signal convergence across operating-point changes. These regions directly guide parameter tuning for heterogeneous GFMs.
arXiv:2603.23640v2 Announce Type: replace Abstract: Deploying large language models on-device for always-on personal agents demands sustained inference from hardware tightly constrained in power, thermal envelope, and memory. We benchmark Qwen 2.5 1.5B (4-bit quantised) across four platforms: a Raspberry Pi 5 with Hailo-10H NPU, a Samsung Galaxy S24 Ultra, an iPhone 16 Pro, and a laptop NVIDIA RTX 4050 GPU. Using a fixed 258-token prompt over 20 warm-condition iterations per device, we measure throughput, latency, power, and thermal behaviour. For mobile platforms, thermal management supersedes peak compute as the primary constraint: the iPhone 16 Pro loses nearly half its throughput within two iterations, and the S24 Ultra suffers a hard OS-enforced GPU frequency floor that terminates inference entirely. On dedicated hardware, distinct constraints dominate: the RTX 4050 is bounded by its battery power ceiling, while the Hailo-10H is limited by on-module memory bandwidth. The RTX 4050 sustains 131.7 tok/s at 34.1 W; the Hailo-10H sustains 6.9 tok/s at under 2 W with near-zero variance, matching the RTX 4050 in energy proportionality at 19x lower throughput. Results should be interpreted as platform-level deployment characterisations for a single model and prompt type, reflecting hardware and software combined, rather than general claims about hardware capability alone.
arXiv:2603.24238v2 Announce Type: replace Abstract: Decentralized cooperative pursuit in cluttered environments is challenging for autonomous aerial swarms, especially under partial and noisy perception. Existing methods often rely on abstracted geometric features or privileged ground-truth states, and therefore sidestep perceptual uncertainty in real-world settings. We propose a decentralized end-to-end multi-agent reinforcement learning (MARL) framework that maps raw LiDAR observations directly to continuous control commands. Central to the framework is the Predictive Spatio-Temporal Observation (PSTO), an egocentric grid representation that aligns obstacle geometry with predictive adversarial intent and teammate motion in a unified, fixed-resolution projection. Built on PSTO, a single decentralized policy enables agents to navigate static obstacles, intercept dynamic targets, and maintain cooperative encirclement. Simulations demonstrate that the proposed method achieves superior capture efficiency and competitive success rates compared to state-of-the-art learning-based approaches relying on privileged obstacle information. Furthermore, the unified policy scales seamlessly across different team sizes without retraining. Finally, fully autonomous outdoor experiments validate the framework on a quadrotor swarm relying on only onboard sensing and computing.
arXiv:2606.00793v2 Announce Type: replace Abstract: Recent advancements in video-based world models have demonstrated an unprecedented ability to synthesize high-fidelity visual sequences. However, a fundamental gap persists between visually plausible video generation and the functional requirements of a world model, particularly in maintaining a stable and reasonable internal state over extended temporal horizons. While existing benchmarks primarily emphasize visual quality, motion coherence, and text-video alignment, they largely overlook memory, the core capability of a world model to preserve consistency across long-term horizons and complex interactions. To address this gap, we present \textbf{MBench}, a comprehensive benchmark dedicated to quantifying and evaluating the memory capability of video world models. We systematically decompose the memory capability of video world models into three hierarchical and complementary core dimensions: entity consistency, environment consistency, and causal consistency, which are further refined into 12 quantifiable sub-dimensions for comprehensive characterization of long-term memory. Our benchmark is built upon rigorously curated real-captured long videos, and evaluated by rule-based quantitative matrices and VLM to enable objective and comprehensive consistency assessment. Extensive evaluations of mainstream state-of-the-art video world models reveal critical systemic limitations of existing methods in long-term state retention, providing a standardized benchmark and clear research direction to advance the field.
arXiv:2606.01304v2 Announce Type: replace Abstract: Hard negative mining has become the dominant strategy for training retrievers, yet it faces intrinsic limitations: negatives are bounded by corpus availability, selected by retriever score rather than diagnostic value, and increasingly contaminated by false positives as the retriever improves. LLM-based synthesis offers a principled alternative, where negatives that are unconstrained, targeted, and free from false positive risk. But we show that naively incorporating generated negatives into contrastive learning often degrades retrieval performance. We identify and formalize the root cause as a generative-discriminative gap: LLM generation optimizes for fluent, plausible text, while contrastive learning demands strategic violations of relevance at the decision boundary. Our analysis reveals two compounding failure modes: discriminative-agnostic generation, where the LLM lacks an explicit model of query information needs and defaults to generic or topic-drifted text that provides no contrastive signal; and source-dependent shortcuts, where distributional artifacts enable the model to distinguish negatives by origin rather than relevance, causing gradient drift that actively corrupts optimization. To close this gap, we propose CausalNeg consisting of two main modules: (1) CoT-guided counterfactual perturbation for data construction: decomposes why a document satisfies a query into explicit information requirements, then surgically violates individual requirements to construct negatives with controlled, interpretable hardness. (2) Query-view entropy maximization during training: disperses generated negatives across the similarity spectrum, minimizing the mutual information between source identity and similarity scores to suppress shortcut exploitation. We make our code publicly available at https://github.com/mzhangzhicheng/CausalNeg.
arXiv:2606.08139v1 Announce Type: cross Abstract: We study charge transport in a magnetized relativistic plasma using kinetic theory within the relaxation-time approximation. By exactly solving the linearized Boltzmann equation in a uniform magnetic field, we obtain an analytic solution for the distribution function in terms of Bessel functions. Using this solution, we compute the full set of retarded current-current correlators and verify the Ward identities. In the hydrodynamic limit, we extract the charge diffusion modes, demonstrating that the transverse diffusion coefficient is strongly suppressed by the magnetic field, scaling as $1/B_0^2$ in the strong-field regime, while the longitudinal diffusion remains unaffected. Furthermore, we analyze the non-hydrodynamic branch cuts in the complex frequency plane, determining their kinematic thresholds and identifying the underlying wave-particle interactions as longitudinal Landau damping and transverse cyclotron damping.
arXiv:2606.08196v1 Announce Type: cross Abstract: We study causal discovery from observational data when some variables are hidden and the data-generating process follows a location-scale noise model (LSNM). Existing methods that handle hidden confounders typically assume additive noise, but in practice, causes often modulate not just the mean but also the variance of their effects. We prove that acyclic directed mixed graphs (ADMGs) satisfying a bow-free condition are identifiable under LSNM with hidden variables, establishing the first identifiability result for causally insufficient models beyond noise additivity. We further provide sufficient conditions for identifying causal direction even when the bow-free assumption is violated. Our two-stage algorithm, LSNM-UV, is sound and complete, and experiments demonstrate improved performance over additive baselines on heteroscedastic data.
arXiv:2606.08505v1 Announce Type: cross Abstract: Speech applications such as meeting transcription and voice agents would benefit from on-device speaker diarization, but practical adoption is limited by inference cost. We study how far a Pyannote 3.1-based pipeline can be accelerated on consumer hardware (an RTX 5070 Ti GPU and an Apple M4 laptop) while preserving diarization error rate (DER). A simple recipe: coarser segmentation stride and per-chunk embedding, yields multi-fold speedups and is DER-neutral on AMI, but degrades sharply on in-the-wild data: on VoxConverse, DER rises from 0.075 to 0.113. We trace the failure to speaker under-counting in the clustering stage, caused by a fixed minimum cluster size interacting with the reduced number of embeddings per speaker. We propose a relative minimum cluster size, mcs = round(f * n) with f = 0.01, which adapts to the embedding budget per recording. A single value of f recovers VoxConverse DER to 0.079 (about 89% of the lost accuracy) while keeping AMI flat, and the accelerated pipeline reaches up to 12.2x speedup on AMI (MPS) over our CAM++ baseline.
arXiv:2602.07345v2 Announce Type: replace Abstract: Distribution Matching Distillation (DMD) is a powerful acceleration paradigm, yet its stability is often compromised in Forbidden Zone, regions where the real teacher provides unreliable guidance while the fake teacher exerts insufficient repulsive force. In this work, we propose a unified optimization framework that reinterprets prior art as implicit strategies to avoid these corrupted regions. Based on this insight, we introduce Adaptive Matching Distillation (AMD), a self-correcting mechanism that utilizes reward proxies to explicitly detect and escape Forbidden Zones. AMD dynamically prioritizes corrective gradients via structural signal decomposition and introduces Repulsive Landscape Sharpening to enforce steep energy barriers against failure mode collapse. Extensive experiments across image and video generation tasks (e.g., SDXL, Wan2.1) and rigorous benchmarks (e.g., VBench, GenEval) demonstrate that AMD significantly enhances sample fidelity and training robustness. For instance, AMD improves the HPSv2 score on SDXL from 30.64 to 31.25, outperforming state-of-the-art baselines. These findings validate that explicitly rectifying optimization trajectories within Forbidden Zones is essential for pushing the performance ceiling of few-step generative models.
arXiv:2602.00056v4 Announce Type: replace Abstract: Large-scale data has fuelled the success of frontier artificial intelligence (AI) models over the past decade. This expansion has relied on sustained efforts by large technology corporations to aggregate and curate internet-scale datasets. In this work, we examine the environmental, social, and economic costs of large-scale data in AI through a sustainability lens. We argue that the field is shifting from building models from data to actively creating data for building models. We characterise this transition as hyper-datafication, which marks a critical juncture for the future of frontier AI and its societal impacts. To quantify and contextualise data-related costs, we analyse approximately 550,000 datasets from the Hugging Face Hub, focusing on dataset growth, storage-related energy consumption and carbon footprint, and societal representation using language data. We complement this analysis with qualitative responses from data workers in Kenya to examine the labour involved, including direct employment by big tech corporations and exposure to graphic content. We further draw on external data sources to substantiate our findings by illustrating the global disparity in data centre infrastructure. Our analyses reveal that hyper-datafication drives substantial and growing environmental costs while systematically redistributing labour risks and representational harms toward the Global South. Thus, we propose Data PROOFS recommendations spanning provenance, resource awareness, ownership, openness, frugality, and standards to mitigate these costs. Our work aims to make visible the often-overlooked costs of data that underpin frontier AI and to stimulate broader debate within the research community and beyond.
arXiv:2602.17245v2 Announce Type: replace Abstract: This position paper argues that building a reliable agentic Web requires shifting from low-level interaction primitives to typed actions supported by a semantic layer. Today's web agents primarily operate through clicks, keystrokes, and DOM manipulation, which leads to brittle long-horizon behavior, high execution cost, and limited auditability. We propose web verbs as a concrete design for this layer. A verb exposes a web operation as a typed function with structured inputs, structured outputs, and documented behavior, whether it is backed by a server-side Web API or a maintained client-side workflow. Verb calls can carry preconditions, postconditions, policy tags, and logging hooks, allowing agents to synthesize concise programs with explicit control flow and data flow and to produce checkable execution traces. Using representative case studies, we illustrate how verb-level composition can produce correct, reproducible outcomes, while browser agents using low-level interaction primitives may produce brittle behavior or incorrect reasoning. We conclude with a call to action on standardization, developer tooling, and community processes needed to make this semantic layer deployable and trustworthy at web scale.
arXiv:2606.08645v1 Announce Type: cross Abstract: We look at Bell inequalities from the lens of information structures in stochastic teams. We consider the usual CHSH game and a dynamic variant of the same to study how various classes of strategies, classical, projective and quantum, behave under team theoretic solution concepts. We find that projective strategies (where each player performs projective measurements) enjoy important properties in the usual CHSH game, but they do not carry over to its dynamic version. These results shed light on the delicate interplay of information structure in quantum strategies and the fragility of some well known ideas under changes of information structure.
arXiv:2602.08222v2 Announce Type: replace Abstract: As post-training optimization becomes central to improving large language models, we observe a persistent saturation bottleneck: once models grow highly confident, further training yields diminishing returns. While existing methods continue to reinforce target predictions, we find that informative supervision signals remain latent in models' own historical weak states. Motivated by this observation, we propose WMSS (Weak Agents Can Make Strong Agents Stronger), a post-training paradigm that leverages weak checkpoints to guide continued optimization. By identifying recoverable learning gaps via entropy dynamics and reinforcing them through compensatory learning, WMSS enables strong agents to improve beyond conventional post-training saturation. Experiments on mathematical reasoning and code generation datasets show that agents trained with our approach achieve effective performance improvements, while incurring zero additional inference cost.
arXiv:2602.08508v2 Announce Type: replace Abstract: The preliminary design of AUGs is intrinsically challenging due to the strong coupling between the external hydrodynamic shape, the hydrostatic balance, the structural integrity, and internal packaging constraints. This complexity is further amplified for bio-inspired configurations, whose rich geometric parametrizations lead to high-dimensional design spaces that are difficult to explore using conventional optimization approaches. This work presents a ML-enabled bi-level multidisciplinary design optimization (MDO) framework for the performance-driven design of a manta-ray-inspired AUG. At the upper level, hydrodynamically efficient external geometries are explored in a reduced design space obtained through physics-driven parametric model embedding, which identifies a low-dimensional latent representation directly correlated with the lift, drag, and pressure distributions. At the lower level, a constrained internal sizing problem determines the minimum feasible empty weight by accounting for structural, hydrostatic, geometric, and payload constraints. To render the resulting bi-level problem computationally tractable, a multi-fidelity surrogate-based optimization strategy is adopted, combining low- and high-fidelity hydrodynamic models with stochastic radial basis function surrogates and adaptive Bayesian sampling. The framework enables efficient exploration of the coupled design space while rigorously managing model uncertainty and computational cost. The optimized configurations exhibit a 14.7\% improvement in maximum hydrodynamic efficiency and a 12.8\% reduction in empty weight relative to the baseline design, while satisfying all disciplinary constraints. These results demonstrate that the integration of physics-driven dimensionality reduction and multi-fidelity machine learning enables scalable and physically consistent MDO of complex bio-inspired underwater vehicles.