arXiv:2606.03788v2 Announce Type: replace Abstract: Sign Language Translation (SLT) is typically evaluated with surface-form metrics such as BLEU and ROUGE, which reward lexical overlap but do not directly measure whether a translation preserves the meaning of the source sign sequence. This is in contrast with the final objective of integrating SLT in assistive technology. In this work, we shift the focus from Sign Language Translation (SLT) to Sign Language Understanding (SLU), with particular emphasis on semantic understanding. Specifically, we evaluate systems based on their ability to correctly recover, from the input video, key semantic aspects of the original sentence, such as actions taking place and facts about people and objects. To enable this evaluation systematically, we propose SLU-2K, a dataset of 2,350 closed-ended video question-answer pairs based on the popular PHOENIX-2014T and CSL-Daily datasets. To obtain SLU-2K, we propose and extensively evaluate an automated data generation pipeline which produces questions across 7 categories, namely actions, locations, numbers, objects, people, time, and weather conditions. We show the potential of SLU-2K by evaluating popular Multimodal Large Language Models (MLLMs) and two representative state-of-the-art systems, MMSTL and SpaMo. Our results show that MLLMs reach near-random performance, highlighting the need for a more systematic integration of SLU in current AI systems. Furthermore, state-of-the-art translation systems carefully fine-tuned on in-domain data still exhibit a substantial semantic gap, with results ranging from 56.7% to 75.2%. These findings suggest that current SLT evaluation protocols overestimate true understanding and that future progress should be measured not only by fluency and n-gram overlap, but also by semantic correctness. Code, prompts, and benchmark files are available at https://github.com/ZenoTsT/SLU-2K
Science Journals
arXiv:2606.04907v2 Announce Type: replace Abstract: Visual navigation requires generating smooth and collision-free trajectories under complex geometric and physical constraints. Existing reactive policies that directly map observations to actions lack anticipatory reasoning, limiting their ability to proactively avoid obstacles. While visual imagination offers predictive foresight, conventional modular approaches separate scene prediction from policy learning, often leading to error accumulation and inefficient inference. To address these limitations, we propose WAM-Nav, a Latent World-Action Model for embodied visual navigation that jointly learns action generation and latent visual foresight, enabling more robust and foresighted navigation decisions without compromising inference efficiency. Specifically, WAM-Nav utilizes a shared Diffusion Transformer for asymmetric joint diffusion to concurrently generate long-horizon actions and short-horizon visual foresight, reducing the inference latency and visual error accumulation inherent in multi-step autoregressive rollouts. To further encourage smooth and consistent trajectory generation, we introduce a dual-stream contextual conditioning mechanism that integrates episode-level ego-motion history with sequential visual observations. Combined with a unified goal alignment module that preserves balanced representations across goal types, WAM-Nav naturally supports Image-Goal, Point-Goal, and No-Goal exploration within a single policy. Extensive experiments on the challenging ClutterScenes and InternScenes benchmarks demonstrate strong generalization of WAM-Nav, particularly on Image-Goal and Point-Goal navigation, where it improves success rates by 15.7% and 3.3%, respectively. Real-world deployment further validates effective zero-shot sim-to-real transfer, achieving an average 85% task success rate across diverse indoor and outdoor environments.
arXiv:2606.03570v3 Announce Type: replace Abstract: Burrows-Wheeler-transform-based compressors rely on local context regularity, but structured text also contains dates, counters, identifiers, coordinates, and other digit runs whose values vary differently from their surrounding tokens. STC is presented as a new algorithm found by the authors through the self-evolving AI system zeelin. It is a practical BWT-family compressor that separates this source of variation before the component BWT stage. It replaces digit runs in the main stream with an unambiguous placeholder and stores the removed digits in length- and context-conditioned side streams. The side streams use stable bucket ordering and compact digit packing, so the decoder can reconstruct the original run order from the normalized main stream without storing a separate permutation. The resulting components are encoded by a fixed internal BWT/M03-style component coder. On enwik9, STC produces a 157,388,188-byte archive with a 183,174-byte decoder source package, giving a local LTCB-style total of 157,571,362 bytes. A full-enwik9 same-coder ablation shows that the digit-context decomposition reduces the archive by 2,629,561 bytes relative to the no-split control. The result is locally verified by full decode and SHA-256 matching; official benchmark status requires independent maintainer-side verification.
arXiv:2509.00836v3 Announce Type: replace Abstract: Motion planning for robotic manipulators is a fundamental problem in robotics. Classical optimization-based methods typically rely on the gradients of signed distance fields (SDFs) to impose collision-avoidance constraints. However, these methods are susceptible to local minima and may fail when the SDF gradients vanish. Recently, Configuration Space Distance Fields (CDFs) have been introduced, which directly model distances in the robot's configuration space. Unlike workspace SDFs, CDFs are differentiable almost everywhere and thus provide reliable gradient information. On the other hand, gradient-free approaches such as Model Predictive Path Integral (MPPI) control leverage long-horizon rollouts to achieve collision avoidance. While effective, these methods are computationally expensive due to the large number of trajectory samples, repeated collision checks, and the difficulty of designing cost functions with heterogeneous physical units. In this paper, we propose a framework that integrates CDFs with MPPI to enable direct navigation in the robot's configuration space. Leveraging CDF gradients, we unify the MPPI cost in joint-space and reduce the horizon to one step, substantially cutting computation while preserving collision avoidance in practice. We demonstrate that our approach achieves nearly 100% success rates in 2D environments and consistently high success rates in challenging 7-DOF Franka manipulator simulations with complex obstacles. Furthermore, our method attains control frequencies exceeding 750 Hz, substantially outperforming both optimization-based and standard MPPI baselines. These results highlight the effectiveness and efficiency of the proposed CDF-MPPI framework for high-dimensional motion planning.
arXiv:2509.14548v2 Announce Type: replace Abstract: High-quality curated datasets are essential for training and evaluating AI approaches, but are often lacking in embodied interactive domains where language and physical action are intertwined. In particular, few datasets capture how people acquire motor skills in embodied tasks through verbal instruction over time. To address this gap, we introduce SimCoachCorpus: a unique dataset of race car simulator driving that enables the investigation of rich phenomena during guided and unguided motor skill acquisition. In this dataset, 29 humans were asked to drive in a driving simulator around a race track for approximately ninety minutes. Fifteen participants received one-on-one instruction from a professional performance driving coach, and 14 participants drove without coaching instruction. SimCoachCorpus includes features such as vehicle state and inputs, map (track boundaries and race-line), and cone landmarks. Additionally, these are synchronized with the coach's concurrent verbal feedback and additional terminal feedback at the end of each lap. We also provide high-quality annotations of high-level coaching categories for each concurrent feedback utterance, ratings on students' compliance with coaching advice, and self-reported cognitive load and emotional state of participants (gathered from surveys during the study). The final dataset includes over 20,000 concurrent feedback utterances, over 400 terminal feedback utterances, and over 40 hours of interactive driving data. Our naturalistic interactive dataset can be used to investigate motor learning dynamics, explore linguistic phenomena, and train computational models of teaching and learning. We demonstrate applications of this dataset for in-context learning, imitation learning, and topic modeling. Data is hosted at https://doi.org/10.7910/DVN/W7VTKZ and code is available at https://github.com/ToyotaResearchInstitute/sim_coach_corpus
arXiv:2605.26290v2 Announce Type: replace Abstract: Temporal signed networks (TSNs) model the time evolution of cooperative and adversarial relationships that arise in applications such as social media analysis, trust and reputation systems, and financial transaction networks. While graph neural networks (GNNs) perform well for static or unsigned link prediction, effective learning in temporal signed graphs remains challenging due to the interaction of signed relations, evolving structure, and balance-theoretic constraints. To address this gap, we propose a \emph{modular} temporal enhancement framework for signed GNNs that integrates historical context into otherwise static architectures. The framework introduces a Historical Context Integration Module (HCIM) that combines learnable recency-aware temporal weighting, LSTM-based embedding trajectory modeling, and multi-head temporal attention to capture both short- and long-term signed interaction dynamics. Historical information is fused with current node representations using either global or node-adaptive weighting, allowing the architecture-agnostic framework to accommodate heterogeneous temporal behaviors. We instantiate the approach on the Self-Explainable Signed Graph Transformer (SE-SGformer), preserving interpretability while extending it with temporal awareness. Experiments on real-world and synthetic TSNs, including Bitcoin OTC, Bitcoin Alpha, Reddit, and small-world network models, demonstrate consistent and statistically significant improvements over the static baseline.
arXiv:2605.27599v2 Announce Type: replace Abstract: Agentic AI workloads - where a single user goal triggers multi-step orchestration, tool calls, retries, and failure recovery - are being targeted for edge deployment, with NVIDIA, Dell, HP, ASUS, MSI, Acer, and Gigabyte all shipping GB10-based desktop AI systems in 2026. We recently demonstrated that orchestration structure dominates agentic energy cost, with workflows consuming 4.33x more energy per successful goal than linear baselines and OOI reaching 7.63x for multi-step reasoning tasks. Separately, Raj et al. show that CPU-side processing accounts for up to 90.6% of total latency and 44% of total dynamic energy in agentic workloads. We report a systematic energy-observability audit of the ASUS Ascent GX10 (GB10 SoC) and find that the platform exposes no CPU energy counter, no INA power-rail monitor, no IPMI/BMC, and no SCMI powercap protocol through any supported software interface. The only on-device energy telemetry is instantaneous GPU power via NVML. We further discover that the MediaTek firmware already computes per-rail energy internally via an undocumented ACPI interface (SPBM), but NVIDIA states there are "no plans to expose CPU rail information." On-device per-process energy attribution - as performed on x86 via RAPL - is therefore not reproducible on this platform through supported interfaces. We formalize a hardware requirements specification for energy-attributed AI, propose an interim calibration bridge for per-domain energy decomposition - confirmed on the Acer Veriton GN100 where CPU energy accumulators are live - and identify a standards-track path via SCMI powercap. Our findings motivate the low-carbon computing community to demand energy observability as a first-class hardware requirement.
arXiv:2606.16537v1 Announce Type: new Abstract: In the classic online stochastic matching proposed by Feldman et al. (FOCS 2009), there is a known bipartite type-graph, where one side of the graph is given offline. Upon the arrival of each online vertex, its type is sampled independently and identically from the other side of the type-graph. This model has been extensively studied over the past decade, yielding a rich body of theoretical results. In this paper, we initiate the study of an edge arrival model for online stochastic matching. In our model, the online edges are sampled independently and identically (KIID) from a known type-graph, which need not be bipartite. We first show that the Greedy algorithm cannot achieve a competitive ratio strictly better than $0.5$ while the Suggested Matching algorithm has a competitive ratio of $1-1/e$ under the assumption of integral arrival rates, matching its performance in the one-sided vertex arrival model. We then propose a two-stage algorithm that combines Greedy and Suggested Matching, and show that its competitive ratio is strictly higher than $1-1/e$ for integral arrival rates. While our algorithm is simple, its analysis is intricate and builds upon the Natural LP, which has been proven very powerful in vertex arrival models. Our result reveals that even in the more challenging edge arrival setting for general graphs, competitive ratios better than $1-1/e$ are still possible, given the known distributions.
arXiv:2606.16233v1 Announce Type: cross Abstract: We show that a radially symmetric, nonresonantly pumped spinor exciton-polariton condensate can acquire chirality without a rotating drive, chiral geometry, or pump orbital angular momentum. Spin relaxation in a two-reservoir system shifts the reservoir-induced blueshift relative to the gain profile, creating an effective non-Hermitian chiral potential. A reduced angular-mode theory reveals tunable exceptional points and non-reciprocal coupling between counter-rotating modes. Full driven-dissipative Gross-Pitaevskii simulations show that this hidden chirality enables all-optical formation of spin-selective half-vortices.
arXiv:2606.16339v1 Announce Type: cross Abstract: Thermodynamic Uncertainty Relations (TURs) impose universal bounds linking current precision to entropy production in nonequilibrium systems. While general bounds like the Proesmans--Van den Broeck (PV) relation provide a broad framework, they often remain loose for processes characterized by renewal events. In this work, we derive a generalized entropic bound for current fluctuations in renewal-reward processes. By utilizing a rigorous variance decomposition within the framework of renewal-reward theory, we obtain a model-independent bound that is not only rigorous but tighter than the standard PV relation. Notably, in the linear-response regime, our bound correctly scales with the renewal rate and identifies a precision penalty arising from cycle-length fluctuations. These results provide a physically informative constraint on the precision of run-and-tumble-type dynamics and highlight the universal limits of transport in complex stochastic walkers.
arXiv:2605.09370v5 Announce Type: replace Abstract: Large-scale AI training is fundamentally a distributed systems problem, where hardware failures are routine operating conditions rather than rare exceptions, yet public operational evidence from production training clusters remains limited. This report presents an empirical analysis of a 63-node NVIDIA B200 production cluster (504 GPUs), using 55 days of Prometheus time-series data and 73 days of operational logs covering 224 multi-node training sessions. The environment is cross-organizational: five parties (SKT, Upstage, Lablup, NVIDIA Korea, VAST Data) share a unified monitoring pipeline. This enabled joint diagnosis of a 60-node-scale storage I/O bottleneck absent in 2-4-node tests, a production-scale phenomenon no single team could isolate alone. We perform three quantitative analyses yielding four findings. First, over 751 Prometheus metrics and 10 XID-identified GPU failures, no single metric is consistently dominant across failure types, motivating multi-signal detection. Second, 523 checkpoint events trace the save/load path from GPU VRAM to the NFS server: restart loading reaches 21.5% of maximum read bandwidth (700 GB/s) and save bursts 16.0% of maximum write bandwidth (250 GB/s), with NFS/RPC queueing and transport-layer backlog rising together. Third, across 224 sessions over 73 days, node exclusions concentrate so the top 3 of 63 nodes account for over 50%. Fourth, auto-retry chain analysis shows a 33.3% success rate over 12 chains (73 attempts), 2.7x the 12.5% manual rate, with a median retry interval of 11 minutes (IQR 10-11). All analyses are grounded in production infrastructure providing session-level workload management, GPU-centric scheduling, and unified observability.
arXiv:2605.25699v2 Announce Type: replace Abstract: We study volume-refined achievability and converse bounds for noisy permutation channels generated by strictly positive DMCs, allowing the reachable output polytope to have arbitrary affine dimension $d\ge 1$. The reachable output polytope may be lower-dimensional than the output simplex, whereas existing refined achievability analyses and fixed-error converses are not adapted to this intrinsic affine geometry. On the achievability side, we develop an affine-coordinate simplex-lattice construction adapted to the reachable output polytope, together with a nearest-neighbor decoder and a geometric error-reduction argument in the same coordinate space. This yields a Gaussian achievability approximation with an $o(1)$ remainder. On the converse side, we first use a meta-converse combined with a KL covering and a local testing estimate to obtain a fixed-error converse with a bounded remainder, which implies the logarithmic $\epsilon$-capacity $d/2$. We then apply the meta-converse with a stratified Jeffreys-mixture auxiliary output distribution. Using a local Laplace approximation and a local likelihood-ratio approximation, this choice identifies the Fisher-volume term and an explicit Gaussian testing constant, yielding a constant-order converse approximation with an $o(1)$ remainder. The achievability and converse constants arise from different constructions and are not claimed to match in general.
arXiv:2603.16273v2 Announce Type: replace Abstract: For field robotic missions such as inspection, search-and-rescue, and exploration, light detection and ranging (LiDAR)-inertial odometry (LIO) can serve as a core component of autonomy by providing localization and mapping in GNSS-denied or unstructured environments. However, transitions between confined and open spaces, which are commonly encountered in field deployments, can induce substantial changes in scan density and local geometric structure, thereby reducing the robustness and computational efficiency of LIO. To address these issues, we present GenZ-LIO, a generalizable LIO framework designed to adapt to variations in spatial scale across confined and open environments. GenZ-LIO comprises three components: (i) scale-aware adaptive voxelization for regulating scan downsampling across spatial scale changes, (ii) hybrid-metric state update for combining point-to-plane and point-to-point residuals under varying geometric structure, and (iii) voxel-pruned correspondence search for efficient point-to-point matching. We conduct a comprehensive evaluation using 42 sequences from nine public datasets and our newly collected NarrowWide dataset to analyze LIO performance under spatial scale variations across diverse field scenarios. Across the evaluated sequences, GenZ-LIO maintains stable odometry estimation without divergence, indicating practical robustness under the tested field conditions. The source code and collected dataset will be made publicly available upon publication.
arXiv:2603.16970v2 Announce Type: replace Abstract: Multimodal egocentric activity recognition integrates visual and inertial cues for robust first-person behavior understanding. However, deploying such systems in open-world environments requires detecting novel activities while continuously learning from non-stationary data streams. Existing methods rely on the main fused logits for novelty scoring, without fully exploiting the complementary evidence available from individual modalities. Because these logits are often dominated by RGB, cues from other modalities, particularly IMU, remain underutilized, and this imbalance worsens as catastrophic forgetting accumulates. To address this, we propose MAND, a modality-aware framework for multimodal egocentric open-world continual learning. At inference, Modality-aware Adaptive Scoring (MoAS) adaptively adjusts modality contributions using sample-wise reliability and refines novelty scoring with deviation and disagreement penalties. During training, Modality-aware Representation Stabilization Training (MoRST) preserves the discriminative capacity of each modality across tasks through modality-specific heads and modality-wise logit distillation. Experiments on a public multimodal egocentric benchmark show that MAND consistently improves novel activity detection and known-class accuracy while substantially reducing FPR95, indicating more reliable open-world recognition. The source code is available at \href{https://github.com/HyeJeongIm/MAND}{github.com/HyeJeongIm/MAND}.
arXiv:2605.26418v3 Announce Type: replace Abstract: A properly calibrated rule-based autoscaler can beat every one of six mainstream deep reinforcement learning (DRL) algorithms on cost across every workload we test - so when, if ever, does DRL actually help? We study this in RLScale-Bench, a reproducible benchmark and evaluation protocol for DRL on adaptive resource control, where an agent allocates compute to a dynamic workload under cost and service-level constraints. We evaluate PPO, DQN, A2C, SAC, TD3, and DDPG under matched architectures, training budgets, and reward functions against a calibrated rule-based baseline across six workload patterns and five seeds (240 runs), instantiate the benchmark on Kubernetes Horizontal Pod Autoscaling, and probe distribution-shift generalization. Three findings challenge common assumptions: (i) the calibrated controller achieves the lowest cost on all six workloads, though it trails the best RL agents on bursty and flash traffic; (ii) discrete-action algorithms outperform continuous-action ones by one to two orders of magnitude in constraint violations due to action-space mismatch; and (iii) no single algorithm dominates across workloads, with rankings shifting by up to four positions. The bottleneck in RL-based resource control is not algorithm selection but baseline calibration, reward engineering, and realistic evaluation protocols.
arXiv:2603.21613v2 Announce Type: replace Abstract: Recommender agents built on Large Language Models offer a promising paradigm for personalized recommendation. However, existing agents typically suffer from a misalignment between their tool-integrated reasoning trajectories and recommendation feedback, limiting their ability to distinguish fine-grained user preferences. To address these challenges, we propose AgenticRec, an agentic recommendation framework that formulates recommendation as a tool-integrated reasoning process over a recommendation-oriented tool suite. Built upon this framework, we further develop a dedicated two-stage training paradigm tailored for recommender agents. In the first stage, we introduce Recommendation-Oriented Trajectory Activation, optimize the agentic recommendation ability under implicit feedback. In the second stage, Progressive Preference Refinement further refines the agent through bidirectional preference reasoning over self-bootstrapped hard pairs, progressively sharpening preference boundaries. Theoretical analysis and extensive experiments demonstrate the effectiveness of AgenticRec. Our code is available at https://anonymous.4open.science/r/AgenticRec-FB16.
arXiv:2605.26595v2 Announce Type: replace Abstract: Large language models (LLMs) are often fine-tuned on uncurated text datasets that adversaries can poison. Existing poisoning attacks primarily rely on fixed trigger phrases that defenses such as outlier detection, clean-data regularization, or online monitoring can neutralize. In this paper, we propose a data poisoning method that teaches an LLM an information hiding scheme reliably and stealthily through semantic associations between shared knowledge such as facts or concepts and attacker-chosen phrases. The induced hiding scheme can encode and decode arbitrary malicious instructions, thus revealing a new and subtle poisoning-induced vulnerability: covert control attacks. We precisely characterize covert control attacks and evaluate them across $5$ LLMs, $3$ backdoor defenses, and $4$ prompt injection defenses. With a small poisoned fraction, covert control attacks outperform heuristic-based prompt injection attacks in average attack success rate by about $40\%$ relative to clean fine-tuned models. They also circumvent defenses based on detection and fine-tuning, maintaining up to $93\%$ attack success rate after backdoor defenses and up to $98\%$ after prompt injection defenses.
arXiv:2606.16563v1 Announce Type: new Abstract: In this paper, we present a structure-preserving rotational discrete gradient method of the full Ericksen-Leslie model for nematic liquid crystal flows with general anisotropic Oseen-Frank elasticity and full Leslie stress. The main difficulty resides in preserving both the pointwise unit-length constraint on the director field and the energy-dissipation at the discrete level, which involves high nonlinearities in the interplay between the length constraint, anisotropic elastic energy, and hydrodynamic couplings. To address this issue, we first reformulate the Ericksen-Leslie system into an equivalent rotational form in which the director evolution is intrinsically tangent to the unit sphere and the coupling terms in the Leslie and Ericksen stresses are reorganized so that the energy exchange and dissipation structure become explicit. Based on this reformulation, we construct rotational discrete gradient schemes that preserve the director length and satisfy an unconditional discrete energy law, together with a fully discrete scheme based on an exact divergence-free spectral approximation. Numerical experiments verify the accuracy, unit-length preservation and energy-stability of the proposed method, and illustrate dynamical effects induced by anisotropic elastic coefficients and shear flow in the full Ericksen-Leslie model.
arXiv:2606.16835v1 Announce Type: cross Abstract: Accurately delineating tissues and reconstructing their chromophore compositions from Multispectral Optoacoustic Tomography (MSOT) images is a key challenge in optoacoustic imaging. The difficulty arises because light fluence distributions within tissue intrinsically depend on spectral optical properties, making the inverse problem inherently ill-posed. Currently, there is a lack of studies leveraging a priori probabilistic anatomical knowledge to guide tissue segmentation and infer chromophore composition. Moreover, most current studies address these two tasks sequentially, which can result in errors accumulating. through the process. To address these issues, we present Anatomical Priors for Reconstruction of Optoacoustic Tomography (APRECOT), a method that leverages probabilistic models of anatomical structures and tissue properties, to enable simultaneous segmentation of tissues and reconstruction of their bulk chromophore compositions. In this proof-of-concept using in-silico data, we show that incorporating probabilistic anatomical context strongly improves the accuracy of bulk chromophore concentration estimation compared to reference methods that do not use any anatomical context or use sequential strategies. This work represents an essential step towards an MSOT imaging mode that directly provides clinically relevant information, such as imaging tissue oxygenation dynamics or disease-related changes in tissue composition.
arXiv:2511.08577v3 Announce Type: replace Abstract: Improving the reasoning abilities of Large Language Models (LLMs), especially under parameter constraints, is crucial for real-world applications. Looped transformers address this by performing multiple latent iterations to refine each token beyond a single forward pass. However, we identify a latent overthinking phenomenon: most token predictions are already correct after the first pass, but are sometimes revised into errors in later iterations. We ask whether selectively skipping latent iterations can improve accuracy, and reveal significant potential with an oracle iteration policy that boosts performance by up to 7.3%. Motivated by this, we propose Think-at-Hard (TaH), a looped transformer optimized for selective iteration. TaH employs a lightweight neural decider to trigger latent iteration, only at tokens likely to be incorrect after the standard forward pass. During latent iterations, depth-aware Low-Rank Adaptation (LoRA) modules shift the objective from general next-token prediction to focused hard-token refinement. A duo-causal attention mechanism extends attention from the token sequence dimension to an additional iteration depth dimension, enabling cross-iteration information flow with full sequential parallelism. Experiments on nine benchmarks show consistent gains across math, QA, and coding tasks. With identical parameter counts, TaH outperforms always-iterate baselines by 3.8-4.4% while skipping iterations on 93% of tokens, and exceeds single-iteration Qwen3 baselines by 3.0-3.8%. When allowing <3% more parameters from LoRA and decider, the gains further increase to 5.3-6.2% and 6.1-6.8%, respectively. Our code is available at https://github.com/thu-nics/TaH.
arXiv:2606.16564v1 Announce Type: new Abstract: Robotic systems routinely encounter conflicting objectives, modeling errors, and degenerate contact conditions that render quadratic programs (QPs) infeasible. Yet most optimization solvers and differentiable QP layers assume feasibility, leading to numerical failures, unstable gradients, or solver breakdown when constraints cannot be simultaneously satisfied. We present Elastic ODYN, a primal--dual non-interior-point QP solver that handles infeasibility through smooth squared-$\ell_2$ elastic relaxations. The resulting formulation remains well posed under ill-conditioning and degeneracy, supports warm starting, and converges to closest-to-feasible solutions when no feasible point exists. A lightweight refinement stage recovers physically meaningful dual variables from the elastic solution. Building on this framework, we develop Elastic OdynLayer, a differentiable QP layer with stable gradients under infeasibility, and Elastic OdynSQP, an infeasibility-aware SQP method that resolves inconsistent subproblems and intrinsically infeasible optimal control tasks through selective constraint relaxation. We evaluate the framework on benchmark QPs, singular contact mechanics, differentiable parameter identification, and quadrupedal and humanoid trajectory optimization. Across all settings, Elastic ODYN consistently outperforms state-of-the-art elastic QP solvers in robustness, warm-start performance, and convergence reliability, enabling optimization, simulation, control, and learning beyond the feasibility assumptions of existing methods.
arXiv:2606.17018v1 Announce Type: cross Abstract: Distributed Acoustic Sensing (DAS) enables the repurposing of existing fiber-optic networks as ultra-dense, long-range seismic arrays for urban monitoring. However, constraints imposed by real-world fiber infrastructure topology and components limit its use for city-scale applications. Recent technological developments have paved the way for short-range, on-chip DAS. Assuming their availability, and based on a Graph Theory framework, we show that monitoring applications fall along a coverage spectrum with two critical thresholds that define three distinct regimes. Low coverage (<10%) can, with optimal design, resolve earthquake early warning, groundwater monitoring, geological mapping, and urban activity tracking. A percolation transition occurs at 51.6% coverage, beyond which the city effectively becomes fully covered and statistical traffic monitoring is possible. Only for effectively complete coverage, infrastructure monitoring, individual vehicle tracking, and pedestrian movement analysis become possible. Thus, privacy-related risks remain very low. We show and exemplify how, for metropolises around the world, an optimal sensing network can be designed for earthquake early warning, traffic monitoring, and urban activity tracking. This framework provides a near-future roadmap for deploying urban DAS networks as a backbone of smart city sensing.
arXiv:2509.21862v3 Announce Type: replace Abstract: How collective behaviors emerge from the interactions of individual LLM-driven agents is a central question in artificial life, yet controlled study of these emergent dynamics has been hindered by the lack of a principled simulation framework for systematic experimentation. To address this, we introduce Shachi, a principled methodology and modular framework that decomposes an agent's cognition into core components: Configuration for intrinsic identity, Memory for contextual continuity, and Tools for extended capabilities, all orchestrated by an LLM reasoning engine. This decomposition treats each cognitive component as an independently controllable variable, enabling perturbation studies that trace how micro-level cognitive traits propagate into population-level dynamics. We investigate behavioral patterns across a 10-task benchmark spanning three levels of collective complexity. Shachi enables memory transfer across environment transitions, producing history-dependent behavioral shifts, and allows agents to simultaneously inhabit multiple environments, revealing cross-environment interference invisible in single-environment studies. Furthermore, in a real-world U.S. tariff shock case study, locally interacting agents with individually controlled cognitive components produce macro-level market dynamics directionally consistent with observed real-world outcomes. Our work provides a rigorous, open-source simulation framework for LLM-based ABM, aimed at fostering cumulative scientific inquiry into the emergent collective behaviors of interacting artificial agents.
arXiv:2603.01696v2 Announce Type: replace Abstract: Large Vision-Language Models (LVLMs) often omit or misrepresent critical visual content in generated image captions. Minimizing such information loss will force LVLMs to focus on image details to generate precise descriptions. However, measuring information loss during modality conversion is inherently challenging due to the modal gap between visual content and text output. In this paper, we argue that the quality of an image caption is positively correlated with the similarity between images retrieved via text search using that caption. Based on this insight, we further propose Cross-modal Identity Mapping (CIM), a reinforcement learning framework that enhances image captioning without requiring additional annotations. Specifically, the method quantitatively evaluates the information loss from two perspectives: Gallery Representation Consistency and Query-gallery Image Relevance. Supervised under these metrics, LVLM minimizes information loss and aims to achieve identity mapping from images to captions. The experimental results demonstrate the superior performance of our method in image captioning, even when compared with Supervised Fine-Tuning. Particularly, on the COCO-LN500 benchmark, CIM achieves a 20% improvement in relation reasoning on Qwen2.5-VL-7B.
arXiv:2603.24071v2 Announce Type: replace-cross Abstract: Although the polarization of soft and tender X-rays is widely used to investigate the diverse physical properties of materials, experimental methods to determine the polarization of tender X-rays (1.5-3.0 keV) remain limited. To address this issue, we propose a method based on the photoelectron angular distribution to detect the polarization of X-rays in this energy range. The angular distribution of photoelectrons emitted from carbon targets was measured using linearly polarized 0.2 to 3.0 keV synchrotron radiation. The photoelectron intensity depends on the angle between the photon's electric field vector and the direction at which the photoelectron is emitted from the target. This result indicates that the photoelectron angular distribution can be used to reliably determine the linear polarization of soft and tender X-rays over a wide range of energy.