arXiv:2606.09356v1 Announce Type: new Abstract: Polymer simulations are among the most computationally demanding workloads in soft-matter research, often requiring days of execution and high energy consumption to achieve physically meaningful results. In this work, we address these challenges through the coupling and optimization of two complementary simulation frameworks: the Uneyama-Doi Model (UDM) and the SOft coarse-grained Monte Carlo Acceleration (SOMA). UDM efficiently propagates concentration fields at the continuum level, while SOMA resolves chain-scale thermal fluctuations via particle-based Monte Carlo dynamics. Each model was individually optimized for GPU execution using kernel fusion, memory coalescing, asynchronous random-number generation yielding up to 70% (UDM) and 80% (SOMA) performance improvement. The coupling is performed through our proposed coordinator library that orchestrates data exchange and synchronizes time-stepping across multiple GPUs. Further management of coupling workload distribution enabled a 13x overall speedup and 24.5x reduction in total energy usage compared to the SOMA baseline, i. e., 96% energy saving. The proposed hybrid approach maintains the same scientific fidelity while drastically reducing the computational and energy footprint, showcasing the potential of energy-aware, cross-application co-design for sustainable high-performance simulations
Science Journals
arXiv:2505.21457v2 Announce Type: replace Abstract: Active vision, also known as active perception, refers to actively selecting where and how to look in order to gather task-relevant information. It is a critical component of efficient perception and decision-making in humans and advanced embodied agents. With the rise of Multimodal Large Language Models (MLLMs) as central planners in robotic systems, the lack of methods for equipping MLLMs with active perception has become a key gap. We first provide a systematic definition of MLLM-based active perception tasks and show that GPT-o3's zoom-in strategy can be viewed as a special case, though it suffers from low efficiency and inaccurate region selection. To address these issues, we propose ACTIVE-o3, a reinforcement learning framework built on GRPO that equips MLLMs with active perception capabilities. Leveraging a modular sensing-action design and a dual-form reward, ACTIVE-o3 autonomously learns efficient and stable region selection strategies without explicit region-selection supervision. We further establish a comprehensive benchmark covering both open-world tasks, including small- and dense-object grounding, and domain-specific scenarios, including remote sensing, autonomous driving, and interactive segmentation. Experimental results demonstrate that ACTIVE-o3 significantly enhances active perception capabilities compared to baselines. Moreover, we show that our framework not only preserves the model's general understanding ability but can also serve as a proxy task for leveraging perception data, further improving performance on benchmarks such as RealWorldQA and MME.
arXiv:2606.09359v1 Announce Type: new Abstract: Integrated modeling of ARC H-modes has been conducted to assess the feasibility of high-performance scenarios compatible with divertor detachment. The analysis incorporates self-consistent evolution of impurity radiation and density profiles, demonstrating that fusion power levels approaching a GW can be achieved while maintaining divertor temperatures below 2 eV with Ar seeding. Sensitivity studies reveal a strong dependence of fusion power on the separatrix density, with performance spanning 750-1000 MW, and a weaker dependence on enrichment factor and pedestal density. Alternative seeding strategies using Neon have also been explored. Plasmas with Argon seeding consistently access H-mode, providing the highest fusion power and detached divertor operation, whereas Neon seeding leads to lower performance (600-850 MW) and less robust H-mode access, due to excessive core impurity accumulation. A small W impurity peaking has been found, with decreasing values at higher Zeff. Further analyses incorporate reduced momentum transport modeling, and sensitivity studies of neoclassical impurity transport, confirming the robustness of the results. Overall, these findings support the viability of high-performing H-mode operation in ARC, ensuring divertor protection, enabled through Argon and Neon impurity seeding.
arXiv:2606.09364v1 Announce Type: new Abstract: Urban populations are increasingly exposed to extreme heat events such as heatwaves, which can be exacerbated in cities due to the urban heat island (UHI) effect. With the aim of developing adaptation strategies, recent years have seen a growing interest in deploying high-resolution measurement networks using low-cost devices (LCDs), which enable the evaluation of intra-urban temperature distribution and its impacts at an unprecedented spatial resolution. However, the reliability of LCD measurements has been called into question, especially regarding potential overheating due to inadequate radiation shielding. In this study, we develop a statistical method to correct temperature biases based on short-wave radiation using a generalized additive model (GAM) and then apply it to LCD measurements in the urban climate networks of the cities of Bern, Lausanne, Neuchatel and Zurich (Switzerland). To that end, we first calibrate the correction procedure to the LCD models used in each city using an intercomparison field study, in which the LCD models are collocated next to a professional automated weather station (AWS) operated by MeteoSwiss in the rural surroundings of Bern. Then, we evaluate how these corrections can influence two climate indices, namely the number of tropical nights and the number of heat warnings issued in each city according to MeteoSwiss heat warning system. The findings suggest that the current AWS underestimate the heat warnings, whereas some LCD models likely overestimate them due to radiative errors. Nevertheless, uncorrected LCD measurements still provide a more reliable estimate of urban temperatures than AWS located outside urban settings. The insights can guide selection of LCD models for new monitoring networks and support the application of model-specific radiative bias corrections to existing LCDs, enabling more accurate assessments of heat and its impacts.
arXiv:2606.09474v1 Announce Type: new Abstract: Generalized Few-Shot Semantic Segmentation (GFSS) has traditionally been approached as a representation-learning problem, requiring task-specific adaptation to incorporate novel classes from limited support examples. Recent foundation models, however, already exhibit strong open-vocabulary recognition and segmentation capabilities, raising a different question: can GFSS be solved through inference-time coordination of frozen semantic priors rather than parameter adaptation? We answer this question with Open-V, a training-free GFSS framework that combines Segment Anything (SAM3) Promptable Concept Segmentation (PCS) with a K-shot CLIP support centroid through calibrated per-pixel semantic arbitration. OpenV introduces no trainable components and supports arbitrary semantic categories at inference time. Beyond segmentation performance, our study contributes three broader findings. First, we show that support information can be incorporated through inference-time semantic grounding, and that its contribution increases as foundation-model text priors weaken on label-disjoint vocabularies. Second, we identify a reproducibility confound in foundationmodel segmentation, demonstrating that preprocessing and evaluation-space mismatches can silently distort reported performance. Finally, we validate Open-V across PASCAL5i, COCO-20i, and ADE-OW, showing that training-free coordination of foundation-model priors generalizes across both conventional GFSS and open-vocabulary evaluation settings. On PASCAL-5i (1-shot), Open-V attains base/novel/harmonic mIoU of 78.4/77.5/77.9, without GFSS-specific training surpassing the strongest trained baseline by +17.7 HM.
arXiv:2503.08434v5 Announce Type: replace Abstract: Recent advances in large-scale text-to-image models have revolutionized creative fields by generating visually captivating outputs from textual prompts; however, while traditional photography offers precise control over camera settings to shape visual aesthetics - such as depth-of-field via aperture - current diffusion models typically rely on prompt engineering to mimic such effects. This approach often results in crude approximations and inadvertently alters the scene content. In this work, we propose Bokeh Diffusion, a scene-consistent bokeh control framework that explicitly conditions a diffusion model on a physical defocus blur parameter. To overcome the scarcity of paired real-world images captured under different camera settings, we introduce a hybrid training pipeline that aligns in-the-wild images with synthetic blur augmentations, providing diverse scenes and subjects as well as supervision to learn the separation of image content from lens blur. Central to our framework is our grounded self-attention mechanism, trained on image pairs with different bokeh levels of the same scene, which enables blur strength to be adjusted in both directions while preserving the underlying scene. Extensive experiments demonstrate that our approach enables flexible, lens-like blur control, supports downstream applications such as real image editing via inversion, and generalizes effectively across both Stable Diffusion and FLUX architectures.
arXiv:2503.14229v4 Announce Type: replace Abstract: Vision-and-Language Navigation (VLN) has been studied mainly in either discrete or continuous spaces, with little attention to dynamic, crowded environments. We present HA-VLN 2.0, a unified benchmark introducing explicit social-awareness constraints. Our contributions are: (i) a standardized task and metrics capturing both goal accuracy and personal-space adherence; (ii) HAPS 2.0 dataset and simulators modeling multi-human interactions, outdoor contexts, and finer language-motion alignment; (iii) benchmarks on 16,844 socially grounded instructions, revealing sharp performance drops of leading agents under human dynamics and partial observability; and (iv) real-world robot experiments validating sim-to-real transfer, with an open leaderboard enabling transparent comparison. Results show that explicit social modeling improves navigation robustness and reduces collisions, underscoring necessity of human-centric approaches. By releasing datasets, simulators, baselines, and protocols, HA-VLN 2.0 provides a strong foundation for safe, human-aware navigation research.
arXiv:2504.00375v2 Announce Type: replace Abstract: The Segment Anything Model 2 (SAM2), a prompt-guided video foundation model, has remarkably performed in video object segmentation, drawing significant attention in the community. Due to the high similarity between camouflaged objects and their surroundings, which makes them difficult to distinguish even by the human eye, the application of SAM2 for automated segmentation in real-world scenarios faces challenges in camouflage perception and reliable prompts generation. To address these issues, we propose CamoSAM2, a motion-appearance prompt inducer (MAPI) and refinement framework to automatically generate and refine prompts for SAM2, enabling high-quality automatic detection and segmentation in VCOD task. Initially, we introduce a prompt inducer that simultaneously integrates motion and appearance cues to detect camouflaged objects, delivering more accurate initial predictions than existing methods. Subsequently, we propose a video-based adaptive multi-prompts refinement (AMPR) strategy tailored for SAM2, aimed at mitigating prompt error in initial coarse masks and further producing good prompts. Specifically, we introduce a novel three-step process to generate reliable prompts by camouflaged object determination, pivotal prompt frame selection, and multi-prompts formation. Extensive experiments conducted on two benchmark datasets demonstrate that our proposed model, CamoSAM2, significantly outperforms existing state-of-the-art methods, achieving increases of 8.0% and 10.1% in mIoU metric. Additionally, our method achieves the fastest inference speed compared to current VCOD models.
arXiv:2504.02983v3 Announce Type: replace Abstract: Metaphor and humor share a lot of common ground, and metaphor is one of the most common humorous mechanisms. This study focuses on the humorous capacity of multimodal metaphors, which has not received due attention in the community. We take inspiration from the Incongruity Theory of humor, the Conceptual Metaphor Theory, and the annotation scheme behind the VU Amsterdam Metaphor Corpus, and developed a novel annotation scheme for humorous multimodal metaphor use in image-caption pairs. We create the Hummus Dataset of Humorous Multimodal Metaphor Use, providing expert annotation on 1k image-caption pairs sampled from the New Yorker Caption Contest corpus. Using the dataset, we test state-of-the-art multimodal large language models (MLLMs) on their ability to detect and understand humorous multimodal metaphor use. Our experiments show that current MLLMs still struggle with processing humorous multimodal metaphors, particularly with regard to integrating visual and textual information. We release our dataset and code at github.com/xiaoyuisrain/humorous-multimodal-metaphor-use.
arXiv:2504.12310v2 Announce Type: replace Abstract: This paper introduces Reflective Empiricism, an extension of empirical science that incorporates subjective perception and consciousness processes as equally valid sources of knowledge. It views reality as an interplay of subjective experience and objective laws, comprehensible only through systematic introspection, bias reflection, and premise-based logical-explorative modeling. This approach overcomes paradigmatic blindness arising from unreflected subjective filters in established paradigms, promoting an adaptable science. Innovations include a method for bias recognition, premise-based models grounded in observed phenomena to unlock new conceptual spaces, and Eureka moments - intuitive insights - as starting points for hypotheses, subsequently tested empirically. The author's self-observation, such as analyzing belief formation, demonstrates its application and transformative power. Rooted in philosophical and scientific-historical references (e.g., Archimedes' intuition, quantum observer effect), Reflective Empiricism connects physics, psychology, and philosophy, enhancing interdisciplinary synthesis and accelerating knowledge creation by leveraging anomalies and subjective depth. It does not seek to replace empirical research but to enrich it, enabling a more holistic approach to phenomena that have not yet been fully grasped. A subsequent body of the author's work is presented as case studies demonstrating the application of the method introduced here.
arXiv:2504.18451v2 Announce Type: replace Abstract: Rapid global population growth underscores the need for digitally enabled agricultural systems that support sustainable food production and data-driven resource management for farmers and stakeholders. The adoption of Internet of Things (IoT) technologies, capable of capturing real-time environmental (e.g., temperature, humidity) and operational (e.g., irrigation) parameters, is a crucial step toward enabling advanced applications such as AI-based yield forecasting. However, the effectiveness of such models is often constrained by limited data availability, particularly in dynamic farm environments where IoT observations must be accumulated over multiple growing seasons. In this study, we deployed IoT sensors in strawberry production polytunnels over two growing seasons to collect data on water usage, internal and external temperature and humidity, soil moisture, soil temperature, and photosynthetically active radiation. These observations were combined with manually recorded yield data spanning four seasons. To address gaps in IoT data for the two seasons without sensor coverage, we developed an AI-based backcasting approach that synthesizes missing sensor observations using historical weather data from a nearby station and existing polytunnel measurements. We then trained AI-based yield forecasting models using both real and synthetic datasets. In this retrospective evaluation, results show that incorporating synthetic data improved yield forecasting accuracy, with models trained on the combined dataset outperforming those using only real sensor, weather, and yield data.
arXiv:2505.07573v2 Announce Type: replace Abstract: Renal mass segmentation has important potential to enhance the clinical workflow, especially in settings requiring quantitative assessments. Kidney volume could serve as an important biomarker for renal diseases, with changes in volume correlating directly with kidney function. Currently, clinical practice often relies on subjective visual assessment for evaluating kidney size and kidney lesions, including tumors and cysts, which are typically staged based on diameter, volume, and anatomical location. To support a more objective and reproducible approach, this research aims to develop a robust, thoroughly validated renal mass segmentation algorithm, named Renal-Net. We employ publicly available training datasets and leverage the state-of-the-art medical image segmentation framework nnU-Net. Validation is conducted using both proprietary and public test datasets, with segmentation performance quantified by Dice coefficient and the 95th percentile Hausdorff distance. Furthermore, we analyze robustness across subgroups based on patient sex, age, CT contrast phases, and tumor histologic subtypes. Our findings demonstrate that our segmentation algorithm, trained exclusively on publicly available data, generalizes effectively to external test sets and outperforms existing state-of-the-art models across all tested datasets. Subgroup analyses reveal consistent high performance, indicating strong robustness and reliability. The developed algorithm and associated code are publicly accessible at https://github.com/DIAGNijmegen/oncology-kidney-abnormality-segmentation.
arXiv:2602.05458v2 Announce Type: replace Abstract: Service-level objective (SLO)-as-code tools make per-service reliability declarative, but users experience journeys: end-to-end executions whose availability and tail latency emerge from topology, routing, redundancy, timeouts/fallbacks, shared failure domains, and tail amplification. Journey objectives are therefore often maintained outside code and drift away from the effective runtime graph. We propose Emergence-as-Code (EmaC), a declarative contract that compiles journey-level SLI bounds and governance artifacts for declared SLOs from intent and evidence. An EmaC specification defines a typed journey expression, leaf bindings to atomic SLOs and telemetry, failure-domain assumptions, and guarded actions. Model Discovery proposes evidence-backed deltas for edges, branch probabilities, redundancy groups, and failure-domain hypotheses; each delta carries provenance and confidence. The compiler derives optimistic and pessimistic journey bounds and emits reviewable governance artifacts. An executable checkout replay shows that local SLOs can remain green while evidence-backed discovery changes the failure-domain model, collapses the pessimistic payment-race bound, and changes the rollout decision from pass to fail or review.
arXiv:2602.08785v2 Announce Type: replace Abstract: Generalization and approximation capabilities of message passing graph neural networks (MPNNs) are often studied by defining a compact metric on a space of input graphs under which MPNNs are equicontinuous. Such analyses are of two varieties: 1) when the metric space includes graphs of unbounded sizes, the theory is only appropriate for dense graphs, and, 2) when studying sparse graphs, the metric space only includes graphs of uniformly bounded size. In this work, we present a unified approach, defining a compact metric on the space of graphs of all sizes, both sparse and dense, under which MPNNs are equicontinuous. This leads to more powerful universal approximation theorems and generalization bounds than previous works. The theory is based on, and extends, a recent approach to graph limit theory called graphop analysis.
arXiv:2602.08916v3 Announce Type: replace Abstract: Objective: Acute mountain sickness (AMS) is the most prevalent altitude illness, affecting unacclimatized individuals ascending above 2,500 m and potentially escalating to life threatening cerebral or pulmonary edema. Conventional machine learning (ML) methods for AMS detection from wearable physiological signals often fail to meet real-time hardware efficiency requirements of continuous monitoring. Methods: We present AMS-HD, the first hyperdimensional computing (HDC)-based framework for real-time AMS detection, spanning high-level bipolar (-1/+1) computing for mobile platforms and low-level binary (0/1) computing for FPGA and ASIC targets. The framework integrates mutual information feature selection, hypervector encoding, and positional projection to enhance classification efficiency. Validation spans ARM, FPGA, and smartwatch-smartphone platforms using wearable-accessible SpO2 and heart rate signals. Results: AMS-HD matches or outperforms SVM and MLP baselines in both binary and multiclass classification, achieving up to 91% accuracy and 90% F1-score in binary classification, and up to 85% accuracy on external AMS-related datasets. On FPGA, AMS-HD reduces LUT and flip-flop usage by 7.3x and 5.8x, while consuming 3.9x less power than MLP. On mobile platforms, AMS-HD requires only 1% battery per session, 60 Bytes of memory, and 2.50 ms inference time -- approximately 2x and more than 3x lower energy consumption than SVM and MLP. Conclusion: AMS-HD provides a scalable, hardware-aware alternative to conventional ML for real-time AMS monitoring, achieving competitive performance with substantially lower resource consumption. Significance: This work presents the first complete HDC framework for altitude sickness detection, bridging wearable inference and low-level hardware deployment for resource-constrained health monitoring.
arXiv:2606.09362v1 Announce Type: new Abstract: Re-Identification (ReID) in autonomous driving is typically formulated as a visual matching problem, where observations of vehicles, pedestrians, and cyclists are associated across time, frames, or camera views using learned appearance embeddings, often complemented by motion, geometric, or multimodal cues. However, purely visual representations may be sensitive to viewpoint, occlusion, illumination, and sensor-domain variations, limiting their interpretability and robustness in complex driving scenes. We propose a baseline study of a zero-shot pipeline using Vision-Language Models (VLMs) to generate textual descriptions of detected traffic participants and evaluate whether these descriptions can support identity matching across observations. Instead of relying only on low-level visual similarity, the proposed formulation represents each object through structured semantic attributes, including category, color, shape, pose, visible parts, spatial context, and distinctive visual cues. This study provides an initial benchmark for language-based re-identification in autonomous-driving scenarios, discussing and evaluating the strengths and limitations of current VLMs for this task. Results demonstrate that zero-shot semantic descriptions can support effective object re-identification, achieving retrieval performance comparable to a supervised CNN baseline while offering greater interpretability through explicit identity cues. However, the experiments also reveal important challenges, including attribute inconsistency across viewpoints and limited fine-grained discrimination between visually similar instances.
arXiv:2606.09416v1 Announce Type: new Abstract: Robot middleware faces a new role in the era of Physical AI. Learned policies, planners, and vision-language-action (VLA) models now enter deployed robots as causal participants on the control path, but the layer that integrates them with timing, scheduling, and network has not been named. Recent language-agent work names this layer the harness, the external system that mediates tools, manages state, bounds resources, and records execution. The robotics community has not yet adopted this framing, and we propose that robot middleware is that harness. A Physical AI harness differs from a software harness in where it intervenes. A software harness mediates at tool-call boundaries. A Physical AI harness must mediate at control, computing, and communication simultaneously, because a learned policy's output crosses all three: its commands shift the trajectory, its inference time shifts the schedule, and its payload shifts the bandwidth. Robot middleware is the lowest robot-stack layer with mediating abstractions over all three, so it is best positioned to compose their enforcement. It already provides most of what a harness needs but lacks the enforcement for an AI model. We name this missing enforcement as three functions: Projection gates each output at emission, Isolation bounds the model's execution and transmission slot, and Transfer falls back to a verified baseline when checks fail. Each appears today as hand-built application code in deployed robot systems, built on surfaces robot middleware already provides. Robot middleware should host them not as the best single-axis enforcer but as the layer that composes all three. We sketch this as a ROS 2 Harness Profile, a deployment artifact that carries an AI model's declared output region, inference budget, and operating regime while the middleware enforces them across ROS 2, DDS, and Zenoh.
arXiv:2606.09378v1 Announce Type: new Abstract: Clinical ultrasound images often contain artificial markers, such as measurement calipers and text, to assist diagnostic interpretation and comparison. However, these markers can introduce shortcut bias in downstream automated analysis, encouraging deep learning models to rely on marker-related cues rather than clinically meaningful anatomy. Existing marker removal methods are either mask-dependent and vulnerable to error propagation, or mask-free deterministic restorers that may over-smooth ultrasound texture and perturb unaffected background regions. To address these challenges, we present Echo-DM, a framework for ultrasound marker removal via conditional latent diffusion and region-aware fusion. Echo-DM follows a common encoder-diffusion-decoder pipeline, where a DiT-based conditional latent diffusion network performs global restoration and a region-aware fusion module enforces preservation-aware image-space refinement under end-to-end mask-free inference. Building on this fixed core design, we further instantiate Echo-DM-V and Echo-DM-R with VAE-based and RAE-based latent modules, respectively, which demonstrates that the Echo-DM architecture is compatible with diverse latent-module instantiations. Extensive experiments on Echo-PAIR, a large-scale paired clinical ultrasound dataset, demonstrate superior marker removal and strong anatomical fidelity compared with representative two-stage baselines, while providing favorable quality--efficiency trade-offs across deployment settings. Data, code and models will be released at https://github.com/MiliLab/Echo-DM.
arXiv:2606.09381v1 Announce Type: new Abstract: Learning robot manipulation policies with deep neural networks from a single demonstration remains highly challenging, as even small deviations from the demonstrated trajectory can quickly compound into failure, while collecting substantial online interaction data is costly. We propose ReGIL, a retrieval-guided imitation learning framework that treats a single demonstration as an external memory. ReGIL repeatedly queries this static memory throughout training to simultaneously guide exploration, generate the regularization buffer, and construct rewards. Specifically, it computes rewards through local temporal alignment between the current trajectory and the retrieved segment, providing step-wise and informative feedback for policy improvement. We evaluate ReGIL on robotic manipulation tasks from the LIBERO and Meta-World benchmarks under the single demonstration setting. ReGIL outperforms prior baselines in both success rate and training efficiency. In real-robot experiments, using only one demonstration and less than one hour of online training, ReGIL achieves over 75% success rate across three manipulation tasks with randomness in both initial robot pose and target position. These results demonstrate that leveraging the single demonstration as reusable memory can provide more than static supervision for efficient robot learning. More details can be found on our website: https://regil2026.github.io/
arXiv:2506.22089v2 Announce Type: replace Abstract: We consider the problem of a game theorist analyzing a game that uses cryptographic protocols. Ideally, a theorist abstracts protocols as ideal, implementation-independent primitives, letting conclusions in the "ideal world" carry over to the "real world." This is crucial, since the game theorist cannot--and should not be expected to--handle full cryptographic complexity. In today's landscape, the rise of distributed ledgers makes a shared language between cryptography and game theory increasingly necessary. The security of cryptographic protocols hinges on two types of assumptions: state-of-the-world (e.g., "factoring is hard") and behavioral (e.g., "honest majority"). We observe that for protocols relying on behavioral assumptions (e.g., ledgers), our goal is unattainable in full generality. For state-of-the-world assumptions, we show that standard solution concepts, e.g., ($\epsilon$-)Nash equilibria, are not robust to transfer from the ideal to the real world. We propose a new solution concept: the pseudo-Nash equilibrium. Informally, a profile $s=(s_1,\dots,s_n)$ is a pseudo-Nash equilibrium if, for any player $i$ and deviation $s'_i$ with higher expected utility, $i$'s utility from $s_i$ is (computationally) indistinguishable from that of $s'_i$. Pseudo-Nash is simpler and more accessible to game theorists than prior notions addressing the mismatch between (asymptotic) cryptography and game theory. We prove that Nash equilibria in games with ideal, unbreakable cryptography correspond to pseudo-Nash equilibria when ideal cryptography is instantiated with real protocols (under state-of-the-world assumptions). Our translation is conceptually simpler and more general: it avoids tuning or restricting utility functions in the ideal game to fit quirks of cryptographic implementations. Thus, pseudo-Nash lets us study game-theoretic and cryptographic aspects separately and seamlessly.
arXiv:2507.08064v4 Announce Type: replace Abstract: As multimedia content expands, the demand for unified multimodal retrieval (UMR) in real-world applications increases. Recent work leverages multimodal large language models (MLLMs) to tackle this task. However, their large parameter size results in high training costs and low inference efficiency. To address this, we propose PUMA: a Layer-Pruned Language Model for Efficient Unified Multimodal Retrieval with Modality-Adaptive Learning. Our approach improves UMR from both structural and learning perspectives. (1) Structurally, we propose Layer-Pruned Self-Distillation, which prunes MLLMs by keeping only shallow layers while distilling features from dropped deep layers as teacher signals. This reduces parameters and preserves representation capability. (2) On the learning side, we introduce Modality-Adaptive Contrastive Learning Loss (MAC-Loss), which separates in-batch negatives into harder intra-modality and easier inter-modality groups based on the target modality, assigning different temperature strategies to enhance learning efficiency. Experiments show our method significantly reduces resource usage while maintaining strong performance.
arXiv:2606.09390v1 Announce Type: new Abstract: Body movement communicates intent at distances and in conditions where neither the face, nor speech can be captured. We study the recognition of communicative intent from 2D body pose alone. We argue that body motion is a reliable signal especially in scenarios that require real time low-cost on-device person-to-robot communication in long distance environments, such as rescue missions. However, existing resources do not isolate this signal. Affective corpora combine body, face, voice and text, while skeleton action-recognition benchmarks label the action performed rather than the message conveyed. We release a dataset of real frames of full-body pose covering ten communicative intents and we compare it against other real (IPC) and synthetic (MotionLCM, VEO3.1, Kimodo) ones that span a range of difficulty. We target systems that can run on a robot's limited onboard hardware. We benchmark multiple models, from skeleton graph classifiers to joint motion-forecasting networks, and report performance metrics together with frame rate on an embedded GPU (NVIDIA Orin~Nano), since speed matters as much as accuracy in our scenario. Finally, we show that a model's own autoregressive self-consistency works as an unsupervised reliability signal. We give a short proof that bounds the probability that a self-consistent prediction is correct, show that this probability grows with the number of consistent steps, and identify the conditions under which a confident prediction can still be false, benchmarked against industry-standard metrics.
CapRL++: Unified Reinforcement Learning with Verifiable Rewards for Dense Image and Video Captioning
arXiv:2606.09393v1 Announce Type: new Abstract: Image and video captioning are fundamental tasks that bridge the visual and linguistic domains, playing a critical role in pre-training Large Vision-Language Models (LVLMs). Current state-of-the-art captioning models are typically trained with Supervised Fine-Tuning (SFT), a paradigm that relies on expensive, non-scalable annotations and often causes models to memorize specific ground-truth answers, limiting their generality and ability to generate diverse, creative descriptions. To overcome these limitations, we propose applying Reinforcement Learning with Verifiable Rewards (RLVR) to the open-ended task of multimodal captioning. We introduce Captioning Reinforcement Learning++ (CapRL++), a novel reference-free training framework that redefines caption quality through its utility: a high-quality caption should enable a non-visual language model to accurately answer questions about the corresponding visual content. CapRL++ employs a decoupled two-stage pipeline where an LVLM generates a caption, and the objective reward is derived from the accuracy of a separate, vision-free LLM answering Multiple-Choice Questions based solely on that caption. Evaluations on more than 20 image and video benchmarks show that CapRL++ improves dense caption quality and strengthens caption-based pretraining across tasks such as spatial and temporal understanding. Pretraining on scalable image and video caption datasets annotated by CapRL++ yields substantial downstream gains. Furthermore, within the Prism Framework for caption quality evaluation, compact models trained with CapRL++ achieve dense captioning performance comparable to substantially larger models such as Qwen2.5-VL-72B and Qwen3-VL-235B-A22B. These results validate that CapRL++ effectively trains models to produce generalizable, high-fidelity descriptions, establishing a robust foundation beyond the limitations of traditional SFT.
arXiv:2606.09407v1 Announce Type: new Abstract: This paper introduces a novel class of delayed functional observers specifically designed to reconstruct delayed control laws under severe output measurement lags, directly complementing recent literature \cite{trinhnn26, trinhnam26}. By systematically mitigating simultaneous, unequal delays across both the actuator and sensor channels, the proposed architecture resolves dual-channel latency without requiring full-state estimation or computationally intensive real-time distributed integration. Ultimately, this work provides a powerful, low-order framework that bridges the gap between idealized control theory and the practical constraints of modern networked engineering systems.
arXiv:2606.09408v1 Announce Type: new Abstract: We present an ethnographic study of an alternative approach to data work, developed by a civic-tech initiative that builds datasets for training and benchmarking online safety systems. They aim to respond to online safety concerns from a feminist perspective, by building safety datasets collaboratively with those most impacted by online harms. In this paper, we examine how this approach aims to reorient data work as a site for repair and redress, and trace the struggles they encounter in the process. Specifically, we draw attention to the challenges and tensions involved in advancing just reward for data work and collective governance of AI datasets. Examining these challenges through an STS-informed lens of reparative justice and repair, we argue that the work of repairing data work (and AI) lies, fundamentally, in resetting the ties of accountability. At a time heightened emphasis on efforts like safety evaluations and red teaming to make AI more responsible, we highlight the need to confront foundational questions about how the humans involved in these efforts relate to the datasets and systems they help produce. A reparative lens demands that we interrupt prevailing norms of data work and place at their centre, not AI or datasets, but those most harmed by the neglect, oversight and exclusion animated in the current modes of dataset production. This, we argue, offers a bold vision for responsibility and contributes towards a critical agenda for building alternative futures of data and AI practice.