arXiv:2605.26362v1 Announce Type: new Abstract: In many reasoning tasks, large language models (LLMs) rely on structured external knowledge, such as graphs and tables, which is typically linearized into sequential token representations. However, even when sufficient knowledge is available, LLMs can still produce hallucinated outputs, and the underlying mechanisms behind such failures remain poorly understood. We investigate these mechanisms and find that hallucinations arise from systematic internal dynamics rather than random noise. First, attention disproportionately concentrates toward shortcut-like structural cues rather than distributing across the full context. Second, feed-forward representations fail to ground the provided knowledge, causing the model to revert to parametric memory. Moreover, our results indicate that hallucination is consistently associated with failures in semantic grounding within feed-forward layers, while attention allocation exhibits greater task-dependent variability. Finally, we show that these mechanistic patterns generalize beyond single-hop graphs to multi-hop and tabular settings, enabling effective hallucination detection across structured knowledge formats.

arXiv:2605.26232v1 Announce Type: new Abstract: Pre-trained video large language models excel at visual reasoning. However, they struggle when videos arrive with auxiliary streams, such as audio, depth map, or dense temporal evidence. In such a scenario, uniform fusion induces modality interference, allowing irrelevant channels to distract the model. To address this issue, we present a unified multimodal video understanding framework, named UniMVU, that performs instruction-aware fusion across video, audio, depth map, or any other modality inputs via two levels of dynamic gating: inner-modality gates emphasize salient regions within each modality, whereas modality-level gates re-weight whole streams; both are conditioned on the text instruction to adaptively balance modality importance. Our UniMVU combines cross-modal self-attention with instruction-driven inner-modality gating module and a modality-level gating module with control token; for time-aligned streams we further adopt a fast-to-slow fusion scheme that reduces redundancy. Across six benchmarks (AVQA, AVSD, Music-AVQA, ScanQA, SQA3D and MVBench), our UniMVU achieves consistent gains over static-fusion baselines achieving gains as high as 13.5 in terms of CIDEr metric. Further, our analysis shows that the gating mechanism aligns with the human-interpretable modality relevance, and ablations show the contributions of inner-modality and modality-level gating. Our UniMVU provides a simple, unified recipe for instruction-aware multimodal video understanding that scales to diverse modalities without hand-crafted fusion rules.

arXiv:2605.26236v1 Announce Type: new Abstract: Co-speech gesture generation requires both semantic expressivity and biomechanically plausible rhythmic motion. Existing holistic gesture models mix lexically grounded semantic gestures with frequent prosody-aligned beat gestures. This limits semantic grounding, speech-motion alignment, and kinematic smoothness. We propose \emph{DuoGesture}, a neuro-inspired and biomechanically informed dual-stream approach that decomposes co-speech gesture synthesis into coupled semantic and beat streams. The two streams are coordinated by a \emph{Semantic Variational Information Bottleneck}, a stochastic frame-level gate that learns when semantic gestures should override rhythmic beat motion. The semantic stream is controlled by \emph{Motion-Grounded Semantic Conditioning}, which replaces purely linguistic word embeddings with motion-language representations to provide motion-aligned semantic priors for long-tailed lexical triggers of gestures. The beat stream is further regularised by an \emph{Inertial Beat Prior}, an anthropometry-weighted arm-chain module that reduces jitter and improves rhythmic consistency without constraining semantic frames. Objective evaluations and subjective experiments show that DuoGesture outperforms strong holistic baselines, while component ablations confirm the complementary roles of semantic grounding, stochastic stream selection, and biomechanical regularisation.

arXiv:2605.26239v1 Announce Type: new Abstract: In this work, we study Cooperative Spatial Intelligence, the ability of decentralized embodied agents to coordinate effectively under dynamic environmental constraints across city-scale outdoor domains. We introduce Sentinel Challenge, a benchmark where multiple decentralized embodied agents must communicate in natural language to agree on a mutually safe and convenient meeting point within large, city-scale outdoor environments. Each agent must then navigate safely while avoiding dynamic sentinels patrolling the area, using a tool that provides coarse spatial information. To address this, we propose CoSaR (Cooperative Spatial Reasoning and Planning), a framework that bridges the high-level communication and planning abilities of foundation models with the precision of classical spatial navigation algorithms. CoSaR enables agents to exchange situational updates, reason over evolving spatial constraints, and collaboratively replan trajectories. Evaluated across 14 city-level scenes with 3-5 agents, CoSaR consistently leads to faster gathering, shorter path lengths, and improved safety. Our results demonstrate that integrating dynamic communication with spatial reasoning is essential for robust multi-agent cooperation. By formalizing this new setting and providing a scalable benchmark, we aim to build a foundation for advancing cooperative spatial intelligence in embodied multi-agent systems. Code and challenge are available at https://github.com/UMass-Embodied-AGI/Sentinel.

arXiv:2605.26241v1 Announce Type: new Abstract: Success in generative modeling across language, image, and video demonstrates that large, well-curated datasets are the key driver for building capable models. 3D Human motion, however, has lagged behind, constrained by an unsatisfying choice between small, high-fidelity motion capture datasets and large-scale in-the-wild collections dominated by static or low-quality sequences. We introduce RoMo, a rich, large-scale, carefully curated dataset of in-the-wild human motions that resolves these tradeoffs. To ensure quality, we introduce a taxonomy-aware filtering pipeline that aggressively removes static and artifact-prone sequences. Every sequence is annotated with detailed captions and organized by a novel three-level semantic taxonomy. This hierarchical structure enables fine-grained, per-category evaluation, that reveals model strengths and weaknesses obscured by global metrics. We demonstrate that models trained on RoMo achieve state-of-the-art fidelity and diversity while gaining a superior understanding of complex, subtle text prompts. Finally, we release the Motion Toolbox to standardize metrics, data conversion, and visualization, establishing a foundation for reproducible and interpretable motion generation research.

arXiv:2605.26244v1 Announce Type: new Abstract: Audio-visual generation is rapidly advancing from short clips to minute-long content, while existing evaluation protocols remain largely confined to short-form settings. Existing benchmarks primarily focus on 5--10 second text-conditioned generation and rarely support unified evaluation across text, image, and video conditioning modalities. Moreover, they provide limited insight into how identity consistency, narrative coherence, and audio-visual alignment degrade over extended temporal horizons. To bridge this gap, we introduce LongAV-Compass, a systematic benchmark for minute-long audio-visual generation. LongAV-Compass contains 284 curated test cases spanning text-to-audio-video (T2AV), image-to-audio-video (I2AV), and video-to-audio-video (V2AV), organized by application scenario and generation complexity. The benchmark combines taxonomy-guided benchmark construction with a unified evaluation framework that integrates MLLM-assisted assessment with complementary perceptual and multimodal metrics, including DINO-v2, ArcFace, CLIP, and ImageBind. The framework evaluates more than 20 fine-grained dimensions covering within-segment quality, cross-segment consistency, global narrative coherence, semantic alignment, and audio-visual synchronization. Through experiments on 11 representative models together with human-alignment validation, LongAV-Compass provides a diagnostic testbed for analyzing the limitations of current systems in sustaining coherent, semantically aligned, and temporally consistent minute-scale audio-visual generation across diverse input modalities.

arXiv:2605.26262v1 Announce Type: new Abstract: Museums are important sites for the dissemination of culture and art. They are institutions rooted in history and tradition; their exhibitions are often designed to highlight these aspects. Recently, a new approach is being explored in the field: emotion-based exhibitions. These exhibitions are designed specifically to elicit emotions in the visitors, in order to maximize engagement, and as a way to democratize access to art and attract a wider, more diverse audience. To do so, the emotional content of the artworks must first be extracted, however, manually annotating the artworks by experts is a prohibitively labor-intensive process, and risks introducing the personal bias of curators. To assist the museum curators in their design of these exhibitions, we wish to develop a tool that can predict the emotional response evoked by a work of art. In this article, we leverage a continuous bi-dimensional emotion space to enhance emotion representations and the training process of deep learning models. Drawing inspiration from existing categorical and dimensional emotion representations, we introduce a new representation, Dimensional Distribution Emotion State (DDES), along with a pipeline for multi-dataset training. We show that DDES provides multiple advantages compared to widely used representations while exhibiting similar baseline performance.

arXiv:2605.26273v1 Announce Type: new Abstract: Semantic segmentation in complex environments such as urban driving scenes remains challenging under adverse lighting conditions, where RGB images alone provide insufficient information. RGB-Thermal fusion leverages the complementary strengths of visible and infrared imagery to improve scene understanding; however, effectively integrating these heterogeneous modalities at varying levels of feature abstraction remains an open problem. In this paper, we propose a multi-modal fusion architecture built upon dual ConvNeXt V2 backbones that employs stage-wise, modality-adaptive fusion strategies. For early-stage features, we introduce a Frequency-Based Fusion Module that decomposes infrared features into low- and high-frequency components via Gaussian filtering, applies dual-branch spatial attention to selectively emphasize thermal patterns and fine-grained boundaries, and integrates them with RGB features through a confidence-gated residual mechanism. For late-stage features, we design a semantic fusion module with cross-modal attention and multi-scale depthwise convolutions to capture semantic correspondences across modalities. The fused features are decoded via a PANet-style bidirectional decoder with deep supervision. Experiments on MFNet and PST900 demonstrate that our lightest variant achieves 61.73\% and 86.24\% mIoU, respectively, with only 35.43M parameters, outperforming recent methods while using substantially fewer parameters and lower computational cost. Code is available at https://github.com/ismailemrecntz/VISIBLE-INFRARED-SENSOR-FUSION

arXiv:2605.26130v1 Announce Type: new Abstract: Operational weather prediction at kilometer scales remains computationally prohibitive for traditional numerical weather prediction (NWP) models, limiting forecast access for applications in energy, agriculture, and disaster management that require fine-grained spatiotemporal detail. Here we introduce AirCast-SR, a foundation model for atmospheric super-resolution that downscales global AI weather forecasts from 0.25 degree (~28 km) to 1 km horizontal resolution at hourly temporal resolution, producing 67-hour forecasts of eight coupled surface variables simultaneously. EarthMind-SR employs a three-dimensional U-Net conditioned within a Latent Consistency Model (LCM) diffusion framework, trained on patch-based samples over the contiguous United States (CONUS) using GraphCast forecasts as input and NOAA's Analysis of Record for Calibration (AORC) as the target. The model achieves near-zero bias across all variables and lead times, and its radial power spectral density analysis demonstrates preservation of fine-scale atmospheric structure at wavelengths of 10 km to 100 km where coarser models lose spectral power. We validate EarthMind-SR across three CONUS case studies spanning winter, summer, and spring seasons, and demonstrate zero-shot global transferability over India and Germany using independent surface station observations without any retraining or fine-tuning. As an open-weights foundation model, EarthMind-SR establishes a new paradigm for kilometer-scale AI weather prediction and provides a platform for regional fine-tuning, distillation, and downstream applications in climate services and hazard forecasting.

arXiv:2605.26121v1 Announce Type: new Abstract: LLM pre-training efficacy increasingly depends on data composition rather than sheer volume. Yet, optimal mixing is hindered by categorization flaws: human taxonomies suffer from ontological misalignment, and Euclidean clustering fails to address embedding anisotropy. We introduce GEM (Geometric Entropy Mixing), a framework reformulating data curation as a variational problem on the hypersphere augmented with a mixing-balance regularizer. By decoupling the generative prior and optimizing the objective via a provable MM (Minorize-Maximize) algorithm, GEM effectively counteracts the cluster collapse to discover balanced semantic structures invisible to Euclidean heuristics. We employ teacher-student distillation to scale this geometric fidelity to web-scale corpora and introduce the Geometric Influence Score (GIS) for interpretable taxonomy generation. Experiments with 1.1B-parameter models demonstrate that GEM establishes a new state-of-the-art when integrated into mixing strategies like DoReMi and RegMix, improving average downstream accuracy by up to 1.2% and offering a robust coordinate system for predictable data mixing.

Yoolim Lee / Bloomberg: Samsung's largest union approves a pay deal that would give chip workers an average bonus of ~$340K, with ~74% of members voting in favor, staving off a strike  —  Samsung Electronics Co.'s largest union voted in favor of a compensation deal that will hand chip workers an average bonus of about $340,000 …

China’s missile-armed robot dogs and humanoid systems point to a new phase in military robotics, raising questions about AI warfare. The post Inside China’s Push to Build an Army of AI-Powered Combat Robots appeared first on TechRepublic.

See how Withings ScanWatch 2 helps users understand glucose spikes, Time in Range, Diabetic Mode, and Lingo-powered blood sugar trends. The post Withings ScanWatch 2 and Glucose Tracking: How It Helps Explain Blood Sugar Spikes appeared first on TechRepublic.

Discussion | Link

Bloomberg: SK Hynix tops $1T in market value after shares jumped 11% on Wednesday, becoming the third Asian company to hit the mark; shares surged 900%+ in the past year  —  SK Hynix Inc. joined Micron Technology Inc. in surpassing the $1 trillion market capitalization milestone, reflecting investors …

Healthcare data lives in dozens of systems, EHRs, claims, labs, pharmacy, SDoH, each...

A Friday memo from the USCIS disrupted weekend plans for several immigration lawyers. Robert Nickelsberg/Getty Images USCIS issued new green-card guidance before the holiday weekend. BI spoke with six immigration lawyers about the fallout from the memo. Lawyers said anxious clients asked whether their yearslong plans had changed overnight. Lynden Melmed was supposed to be spending part of his Friday with family visiting from Germany. They were touring the Capitol and learning how the American government works during the long holiday weekend. Then, the US Citizenship and Immigration Services released a memo about green card applicants. Melmed, a partner at BAL and former US Citizenship and Immigration Services chief counsel, stayed behind to read it and field client questions. "You do unfortunately need to clear your schedule," Melmed said. "That's just an occupational hazard of being an immigration lawyer." On Friday, USCIS announced that it would grant "adjustment of status" — the process that lets some immigrants in the US apply for green cards without leaving — only in "extraordinary circumstances." The decision could force many immigrants to leave the country and continue their green card applications abroad rather than completing the process in the US. A USCIS spokesperson, Zach Kahler, said on Friday that the new guidance likely wouldn't impact "people who present applications that provide an economic benefit or otherwise are in the national interest." Business Insider spoke with six immigration lawyers across the US who work with tech workers, startup founders, physicians, investors, and other foreign national workers. They described a holiday weekend packed with calls, texts, and emails from anxious clients trying to understand whether yearslong green-card plans had changed overnight. The questions were practical: Should workers keep filing green-card applications? Should they wait? Would pending cases be affected? Should people avoid international travel? Companies were also asking whether this was serious enough to brief senior executives immediately. For now, the answer was a cautious wait-and-see. "I started hearing from my clients and from other immigration attorneys within minutes of this memo dropping on Friday morning," said Loren Locke, an immigration attorney who works with multinational corporate clients. "It has thrown a lot of uncertainty into something that's been very stable and very predictable for decades, out of nowhere, with no warning." Brian Hunt, counsel at immigration firm Fragomen, said his company began hearing from clients on Friday and "pretty much worked all weekend." "Everyone wants answers as to what this memo means," he said. For employers, the concern is not abstract. Consular processing can be slow and unpredictable, lawyers said, and companies may struggle if workers have to leave the US without knowing when they can return. "I don't know how people could just leave their job for months and come back," Hunt said. Multiple lawyers compared the rollout to a September presidential proclamation signed by President Donald Trump that raised the H-1B petition fee to $100,000, which sparked immediate alarm before later guidance softened its apparent impact. Several attorneys also said Friday's announcement appeared more severe than the underlying memo. At Bay Immigration Law, which works with startup founders and tech workers, Otto Van Maerssen said many existing clients were seeking reassurance. "For some of them, it was, is it even possible now to adjust status?" Van Maerssen, a senior partner, said. TJ Albrecht, managing director at Bay Immigration, estimated that client outreach surged over the long weekend. He said the firm's reaction oscillated between "dread and optimism" as lawyers compared the memo with the USCIS press release announcing the change. "So, we think that the vast majority, at least from our clients, will ultimately not be affected," he said. Other visa applicants — like students and B-1 temporary business visitors — might not be so lucky. Divij Kishore, founder of the immigration-focused firm Flagship Law, said clients asked whether they should continue with green-card applications, what would happen to pending cases, and whether staying in the US still made sense. "There's a sense of fatigue that I'm starting to see in the people that I represent," Kishore said. "I'm concerned that in the way that it's been released to the public and the way it's been reported to the public so far, there's a knee-jerk reaction that's happening where people are acting out of fear rather than proactive decision-making and thoughtful decision-making." Locke said the memo arrived at the end of a yearslong process for some workers who had followed the rules, renewed visas, built careers, and started families in the US. "It has been chaotic," she said. "Right now, we are waiting to see what USCIS does." USCIS didn't immediately respond to a request for comment from Business Insider. Read the original article on Business Insider

Rebecca Torrence / Bloomberg: Sources: Fireworks AI, which helps companies run AI models, is in talks to raise funding at a $15B valuation after being valued at $4B in October 2025  —  Fireworks AI, a startup that helps companies run artificial intelligence models, is in talks to raise a new round of funding that would value …

Ahead of global elections, we’re helping people access information, supporting cyber defenders, and increasing AI transparency