arXiv:2606.16887v1 Announce Type: new Abstract: Integrated nanomechanical circuits require compact and predictable ways to read out, confine, and connect mechanical motion across multiple nanoscale elements. This challenge is particularly acute for megahertz flexural modes, whose large mechanical response and nonlinear dynamics are attractive for optomechanics, sensing, and signal processing, but whose extended nature makes local confinement and coupling difficult within dense devices. Here we demonstrate an optomechanical nanobeam platform in which optical transduction and mechanical connectivity are both engineered lithographically. Transverse geometric asymmetry in the photonic-crystal cavity breaks the cancellation that suppresses dispersive coupling to in-plane flexural motion, making these modes optically bright without ancillary structures. In parallel, serpentine mechanical links engineered through their complex band structure act as compact mirrors and evanescent couplers for MHz flexural waves. In coupled-cavity devices, the normal-mode splitting decays exponentially with the number of serpentine cells, yielding an experimental attenuation constant in quantitative agreement with full-system simulations. Geometry-dependent measurements further show that the coupling can be tuned by the interconnect design and identify regimes where finite-link modes hybridize with the cavity modes, beyond a simple two-resonator picture. These results establish complex-band-engineered mechanical links as calibrated interconnects for scalable optomechanical nanocircuits based on optically addressable MHz flexural resonators.
Science Journals
arXiv:2606.16988v1 Announce Type: new Abstract: Benchmark scores tell you what an agent got right; they do not tell you how it got there. In this work, we introduce methods for comparing agents procedurally in different contexts, where the model, tasks, and approaches vary. We compare ten agents and find that they are identifiable by their behavioral habits, which we define as fingerprints: a probe over these procedural signatures attributes an unseen trajectory to the correct agent at 85.7% accuracy, controlling for leakage across tasks. We develop procedural representations for agent problem-solving procedures with an emergent vocabulary induction technique that is meant to be maximally compressive to avoid surface-level variation while being expressive enough to unveil the quirks of the models' patterns. We apply our framework to the software engineering evaluation dataset SWE-Bench to study the structural distinctness of agent trajectories and find that behavior is most similar between models from similar release periods and those that are distilled from one another (e.g., a distilled student model and its teacher have a Jensen-Shannon divergence of 0.25, about half the distance between other model pairs). As more models saturate evaluations, we believe that it will be important to probe model behavior along more holistic dimensions than success rates alone. We introduce ProcGrep, a library for auditing and evaluating agents for how they approach tasks at a procedural level given their traces in a top-down fashion. We believe this work has a range of applications to help developers work with and program coding agents, such as task-aware model routing, agent monitoring, and finer-grained cost analysis.
arXiv:2509.23014v2 Announce Type: replace Abstract: With the powerful reasoning capabilities of large language models (LLMs) and vision-language models (VLMs), many recent works have explored using them for decision-making. However, most of these approaches rely solely on language-based reasoning, which limits their ability to reason and make informed decisions. Recently, a promising new direction has emerged with unified multimodal models (UMMs), which support both multimodal inputs and outputs. We believe such models have greater potential for decision-making by enabling reasoning through generated visual content. To this end, we propose Uni-Plan, a planning framework built on UMMs. Within this framework, a single model simultaneously serves as the policy, dynamics model, and value function. In addition, to avoid hallucinations in dynamics predictions, we present a novel approach self-discriminated filtering, where the generative model serves as a self-discriminator to filter out invalid dynamics predictions. Experiments on embodied decision-making tasks show that Uni-Plan substantially improves success rates compared to VLM-based methods, while also showing strong data scalability, requiring no expert demonstrations and achieving better performance under the same training-data size. This work lays a foundation for future research in reasoning and decision-making with UMMs.
arXiv:2603.02668v2 Announce Type: replace Abstract: We present SorryDB, a dynamically-updating benchmark of open Lean tasks drawn from 78 real world formalization projects on GitHub. Unlike existing static benchmarks, often composed of competition problems, hillclimbing the SorryDB benchmark will yield tools that are aligned to the community needs, more usable by mathematicians, and more capable of understanding complex dependencies. Moreover, by providing a continuously updated stream of tasks, SorryDB mitigates test-set contamination and offers a robust metric for an agent's ability to contribute to novel formal mathematics projects. We evaluate a collection of approaches, including generalist large language models, agentic approaches, and specialized symbolic provers, over a selected snapshot of 1000 tasks from SorryDB. We show that current approaches are complementary: even though an agentic approach based on Gemini Flash is the most performant, it is not strictly better than other off-the-shelf large-language models, specialized provers, or even a curated list of Lean tactics.
arXiv:2603.07633v3 Announce Type: replace Abstract: This paper studies opinion dynamics in multilayer networks. Extending a single-layer model, we formulate opinion updates as a synchronous coordination game in which agents minimize a local cost to stay close to their neighbors' opinions. We propose two coupling mechanisms between layers: (i) a merged model that aggregates layers through weighted influences, and (ii) a switching model that periodically alternates across layers. Using random-walk and spectral analysis, we derive sufficient conditions for consensus, characterize convergence rates, and analyze stability under network perturbations. We show that multilayer interactions can induce or accelerate global consensus even when no single layer achieves it alone, and conversely, that individually coordinated layers may lose consensus once interconnected. A common thread arising from our analysis is that the alignment in the weighted degrees of the nodes between the two layers is the main determinant of whether merging or switching can speed up convergence to consensus compared to layers operating in isolation, providing network design intervention guidelines.
arXiv:2603.04239v2 Announce Type: replace Abstract: Recent breakthroughs in Diffusion Transformers (DiTs) have revolutionized the field of visual synthesis due to their superior scalability. To facilitate DiTs' capability of capturing meaningful internal representations, recent works such as REPA incorporate external pretrained encoders for representation alignment. However, the underlying mechanisms governing representation learning within DiTs are not well understood. To this end, we first systematically investigate the representation dynamics of DiTs. Through analyzing the evolution and influence of internal representations under various settings, we reveal that representation diversity across blocks is a crucial factor for effective learning. Based on this key insight, we propose DiverseDiT, a novel framework that explicitly promotes representation diversity. DiverseDiT incorporates long residual connections to diversify input representations across blocks and a representation diversity loss to encourage blocks to learn distinct features. Extensive experiments on ImageNet 256x256 and 512x512 demonstrate that our DiverseDiT yields consistent performance gains and convergence acceleration when applied to different backbones with various sizes, even when tested on the challenging one-step generation setting. Furthermore, we show that DiverseDiT is complementary to existing representation learning techniques, leading to further performance gains. Our work provides valuable insights into the representation learning dynamics of DiTs and offers a practical approach for enhancing their performance.
arXiv:2602.24012v2 Announce Type: replace Abstract: Contrastive learning has become a cornerstone of modern representation learning, allowing training with massive unlabeled data for both task-specific and general (foundation) models. A prototypical loss in contrastive training is InfoNCE and its variants. In this work, we show that the InfoNCE objective induces Gaussian structure in representations that emerge from contrastive training. We establish this result in two complementary regimes. First, we show that under certain alignment and concentration assumptions, projections of the high-dimensional representation asymptotically approach a multivariate Gaussian distribution. Next, under less strict assumptions, we show that adding a small asymptotically vanishing regularization term that promotes low feature norm and high feature entropy leads to similar asymptotic results. We support our analysis with experiments on synthetic and CIFAR-10 datasets across multiple encoder architectures and sizes, demonstrating consistent Gaussian behavior. This perspective provides a principled explanation for commonly observed Gaussianity in contrastive representations. The resulting Gaussian model enables principled analytical treatment of learned representations and is expected to support a wide range of applications in contrastive learning.
arXiv:2603.05299v2 Announce Type: replace Abstract: Large language models show that simple autoregressive training can yield scalable and coherent generation, but extending this paradigm to speech remains challenging due to the entanglement of semantic and acoustic information. Most existing speech language models rely on text supervision, hierarchical token streams, or complex hybrid architectures, departing from the single-stream generative pretraining paradigm that has proven effective in text. In this work, we introduce WavSLM, a speech language model trained by quantizing and distilling self-supervised WavLM representations into a single codebook and optimizing an autoregressive next-chunk prediction objective. WavSLM jointly models semantic and acoustic information within a single token stream without text supervision or text pretraining. Despite its simplicity, it achieves competitive performance on consistency benchmarks and speech generation while using fewer parameters, less training data, and supporting streaming inference.
arXiv:2603.05373v3 Announce Type: replace Abstract: Neural codec language models enable high-quality discrete speech synthesis, yet their inference remains vulnerable to token-level artifacts and distributional drift that degrade perceptual realism. Rather than relying on preference optimization or retraining, we propose MSpoof-TTS, a training-free inference framework that improves zero-shot synthesis through multi-resolution spoof guidance. We introduce a Multi-Resolution Token-based Spoof Detection framework that evaluates codec sequences at different temporal granularities to detect locally inconsistent or unnatural patterns. We then integrate the spoof detectors into a hierarchical decoding strategy, progressively pruning low-quality candidates and re-ranking hypotheses. This discriminator-guided generation enhances robustness without modifying model parameters. Experiments validate the effectiveness of our framework for robust and high-quality codec-based speech generation. Audio samples and code are available.
arXiv:2603.05876v2 Announce Type: replace Abstract: The generation of synthetic novel views has the potential to positively impact robot navigation in several ways. In image-based navigation, a novel overhead view generated from a scene taken by a ground robot could be used to guide an aerial robot to that location. In Video Place Recognition (VPR), novel views of ground locations from the air can be added that enable a UAV to identify places seen by the ground robot, and similarly, overhead views can be used to generate novel ground views. This paper presents a systematic evaluation of synthetic novel views in VPR using five public VPR image databases and seven typical image similarity methods. We show that for small synthetic additions, novel views improve VPR recognition statistics. We find that for larger additions, the magnitude of viewpoint change is less important than the number of views added and the type of imagery in the dataset.
arXiv:2606.16996v1 Announce Type: new Abstract: Segment Anything Model 3 (SAM 3) provides a strong frozen backbone for concept-prompted segmentation, but applying it directly to open-vocabulary semantic segmentation (OVSS) is inefficient: full-resolution decoding is typically run over the entire dataset vocabulary, whereas each image contains only a small active subset of classes. We introduce ActiveSAM, a training-free, zero-shot inference framework that turns SAM 3 into an active-vocabulary segmenter. ActiveSAM first canonicalizes and expands class prompts, then estimates an image-conditioned active set from a low-resolution presence preview. Only the retained classes are decoded at full resolution, using bucketed prompt multiplexing with the frozen SAM 3 decoder. The preview stage uses only class-presence evidence and skips unnecessary segmentation-head computation, while the final stage applies margin-aware background calibration to suppress low-confidence pixels. ActiveSAM requires no target-dataset training, no weight updates, and no oracle class-presence labels. Across eight OVSS benchmarks, ActiveSAM improves the speed-accuracy tradeoff of training-free open-vocabulary semantic segmentation, outperforming the current state-of-the-art SegEarth-OV3 by approximately +1.4 mIoU on average while running up to 5.5x faster on large-vocabulary datasets. ActiveSAM also demonstrates the strongest robustness under image corruption that simulates real-world distribution shift, making it well-suited for deployment in noisy-input domains such as autonomous driving and embodied AI. Code is available at https://github.com/VILA-Lab/ActiveSAM.
arXiv:2606.17001v1 Announce Type: new Abstract: Cyber-physical system (CPS) controllers are vulnerable to faults and malicious attacks, including failures triggered only under complex plant conditions, yet pre-deployment validation typically relies on plant models or digital twins that exercise the controller's I/O as a black box. Side-channels, used to detect those run-time behavioral anomalies, are complementary but also open-loop, detached from I/O instrumentation, and driven by synthetic inputs rather than realistic plant feedback. We present a closed-loop digital twin framework that bridges this gap by hosting an unmodified controller binary within the SaMOSA Linux sandbox with its I/O rerouted to an external plant simulator, allowing coupled capture of simulated plant conditions and events alongside the controller's behavioral side-channels. The framework captures four time-synchronized controller side-channels (hardware performance counters, system calls, disk activity, and network activity) alongside plant state, and uses orchestration hooks for repeatable, parameterized runs. We demonstrate the framework on an OpenPLC runtime executing a Structured Text control program against a Modbus-connected IEEE 14-bus power-system model, and also discuss the application to robotics systems. The captured side-channels correlate controller behavior with simulated plant events, establishing an observability foundation for online testing, coverage analysis, and anomaly detection in CPS controllers.
arXiv:2601.15148v2 Announce Type: replace Abstract: We consider a game-theoretic variant of an interval scheduling problem. Every job is associated with a length, a weight, and a color. Each player controls all the jobs of a specific color, and needs to decide on a processing interval for each of its jobs. Jobs of the same color can be processed simultaneously by the machine. A job is covered if the machine is configured to its color during its whole processing interval. The goal of the machine is to maximize the sum of weights of all covered jobs, and the goal of each player is to place its jobs such that the sum of weights of covered jobs from its color is maximized. The study of this game is motivated by several applications like antenna scheduling for wireless networks. We first show that given a strategy profile of the players, the machine scheduling problem can be solved in polynomial time. We then study the game from the players' point of view. We analyze the existence of Nash equilibria, its computation, and inefficiency. We distinguish between instances of the classical interval scheduling problem, in which every player controls a single job, and instances in which color sets may include multiple jobs.
arXiv:2606.13053v2 Announce Type: replace Abstract: Pretrained-feature world models provide a useful substrate for robot imagination, but visual or latent prediction alone does not determine whether an imagined future satisfies task-relevant predicates. Long-horizon manipulation requires progress signals that are relational, predicate-level, and physically grounded: whether an object has moved, whether a drawer or contact state has changed, whether a placement predicate is satisfied, and whether a candidate future is reliable enough for execution. We introduce \textbf{EV-WM}, a predicate-grounded verification framework for world-model planning. EV-WM rolls out candidate futures in pretrained visual-feature space, decodes them into structured event states, and scores them using task-progress, semantic-consistency, physical-feasibility, and uncertainty terms. The verifier guides sampling-based planning, gates candidate actions, and, in the contact-sensitive LIBERO wine-rack setting, selects among PPO-generated proposals. Across navigation, deformable-object, wall-constrained, and language-described manipulation studies, EV-WM shows that predicate-grounded verification can make feature-space world-model planning more interpretable and better aligned with task progress.
arXiv:2606.17010v1 Announce Type: new Abstract: Heterogeneous Treatment Effect (HTE) identification is crucial to explain the impact of an intervention and optimize our policies accordingly. Existing approaches trade expressivity for interpretability, but, if some active heterogeneity drivers are unmeasured, methods at both ends of this spectrum allow for spurious HTE characterization with no causal reading. In this work, we focus on controlled experiments and argue that an oracle HTE causal characterization via the latent interactors is now within reach, thanks to (i) more extensive pre-treatment measurements, i.e., multi-modal and multi-view, and (ii) scalable representations with minimal human supervision. We then re-frame HTE identification as a Markov-blanket discovery problem on a sufficient and aligned pre-treatment representation, and introduce Neural EXposure Interaction Search (NEXIS), an iterative procedure with provable and empirically validated consistent selection. We deploy NEXIS on two anti-poverty programs in Africa, augmenting each with satellite imagery capturing previously unmeasured environmental effect modifiers, leading to novel, interpretable and prescriptive guidelines to optimize the programs' next iterations.
arXiv:2606.17011v1 Announce Type: new Abstract: Human interventions provide crucial corrective signals for post-training Vision-Language-Action (VLA) models. However, enabling seamless humanoid interventions is a formidable systems challenge due to complex whole-body kinematics and dexterous-hand control. Consequently, the collected intervention trajectories are often suboptimal, and methods that rely on human interventions as expert supervision can absorb hesitant, inefficient, or even erroneous behaviors. To address both the system and algorithmic challenges, we propose ROVE, a reinforcement learning framework for humanoid VLA post-training with imperfect human interventions. First, ROVE introduces a human-in-the-loop pipeline capable of collecting deployment and intervention data for humanoid manipulation. Second, it utilizes Optimistic Value Estimation (OVE) to prioritize high-value behaviors from mixed-quality trajectories. To further robustify value estimation, we incorporate cross-embodiment human experience videos to provide rich supervision for long-tailed failure and recovery modes. The resulting critic yields informative advantage signals, steering the VLA actor to focus on high-value behaviors rather than indiscriminately imitating all actions. On challenging real-world contact-rich and fine-grained humanoid manipulation tasks, ROVE outperforms experience-learning baselines and consistently improves across multiple rollout-intervention iterations.
arXiv:2601.15459v2 Announce Type: replace Abstract: This study presents an integrated framework for enhancing the safety and operational efficiency of robotic arms in laparoscopic surgery by addressing minimum distance estimation between multi-arm manipulators and the associated collision-aware warning. By combining analytical modeling, real time simulation, and machine learning, the framework offers a robust solution for ensuring safe robotic operations. An analytical model was developed to estimate the minimum distances between robotic arms based on their joint configurations, offering theoretical calculations that serve as both a validation tool and a benchmark. To complement this, a 3D simulation environment was created to model two 7 DOF Kinova robotic arms (Kinova inc., Boisbriand, QC, Canada), generating a diverse dataset of configurations for distance estimation and collision warning. Using these insights, a deep residual neural network model was trained with joint configurations as inputs. On the held out validation set, the model achieves R2 = 0.940, RMSE = 42.0 mm, MAE = 28.7 mm, and a near zero mean bias, demonstrating strong predictive accuracy and consistent generalization across the workspace. The framework is intended as an early collision warning layer, where a warning is triggered when the predicted inter-arm distance falls below a 0.2 m threshold, which corresponds to a surface to surface clearance of approximately 50 mm given the Kinova Gen3 (Kinova inc., Boisbriand, QC, Canada) cross sectional radius. This work demonstrates the effectiveness of combining analytical modeling with machine learning to enhance the precision and reliability of multi-arm robotic systems.
arXiv:2601.16093v2 Announce Type: replace Abstract: Pixel-wise capabilities are essential for building interactive intelligent systems. However, pixel-wise multi-modal LLMs (MLLMs) remain difficult to scale due to complex region-level encoders, specialized segmentation decoders, and incompatible training objectives. To address these challenges, we present SAMTok, a discrete mask tokenizer that converts any region mask into two special tokens and reconstructs the mask using these tokens with high fidelity. By treating masks as new language tokens, SAMTok enables base MLLMs (such as the QwenVL series) to learn pixel-wise capabilities through standard next-token prediction and simple reinforcement learning, without architectural modifications and specialized loss design. SAMTok builds on SAM2 and is trained on 209M diverse masks using a mask encoder and residual vector quantizer to produce discrete, compact, and information-rich tokens. With 5M SAMTok-formatted mask understanding and generation data samples, QwenVL-SAMTok attains state-of-the-art or comparable results on region captioning, region VQA, grounded conversation, referring segmentation, scene graph parsing, and multi-round interactive segmentation. We further introduce a textual answer-matching reward that enables efficient reinforcement learning for mask generation, delivering substantial improvements on GRES and GCG benchmarks. Our results demonstrate a scalable and straightforward paradigm for equipping MLLMs with strong pixel-wise capabilities. Our code and models are available.
arXiv:2606.13408v2 Announce Type: replace Abstract: The 2022--2026 burst of activity in small-format matrix multiplication (AlphaTensor 2022, AlphaEvolve 2025, Schwartz--Zwecher 2025) has produced striking individual results but scattered them across different fields, attribution conventions, and serialisation formats. A complementary line of work -- Perminov's open-source flip-graph framework~\cite{perminov2026fast,perminov2025fast} -- instead drives existing construction methods, notably flip-graph and \emph{meta-flip-graph} search, at scale across large format spaces, discovering many new low-rank schemes (including ternary-integer ones) that further enrich the landscape this catalog must unify. We present a unified, machine-checkable catalog covering shapes up to \nmpshape{32}{32}{32} over \Rationals, \Integers, \Reals, \Complex, and \Ftwo, with a separate axis for commutative algorithms (Waksman 1970, Makarov 1986, Rosowski 2019). Derivation over this catalog is performed by a \emph{frontier-closure search} that recombines catalog entries by axis-flip, Kronecker, axis concatenation, serendipitous products, recombination-with-allocation (with optional output peeling and pair fusion), and downward projection. A central methodological point is the \emph{non-overlap property}: our recombination does not, and cannot, rediscover the shared bilinear products that hand-crafted constructions (Strassen, Laderman, Smirnov, AlphaTensor) are built around. This draws a clean line between the ``find a cleverer bilinear core'' and ``compose known cores'' axes of progress, and resolves several attribution puzzles in the literature. We refresh the DIS09 comparison tables, split per field and with a commutative column, and provide the tooling to regenerate them automatically as the catalog evolves.
arXiv:2606.11349v2 Announce Type: replace Abstract: In hierarchical reasoning, failures often originate at intermediate decision points where the agent commits to a wrong branch without recognizing that it lacks critical information. Rather than treating clarification as an external uncertainty trigger, we propose ACTION-RATING, a formulation that places it inside the agent's action space on a shared ordinal scale with navigation, so that asking competes directly with acting at every decision point and help-seeking becomes observable at intermediate states. Two structurally distinct information-seeking modes emerge from the agent's own ratings: mandatory (no viable branch) and opportunistic (residual uncertainty despite a leading candidate). On Harmonized Tariff Schedule classification (30,000-node taxonomy, three benchmarks, 9~LLMs across 4 families), we observe a regime shift from mandatory to opportunistic clarification, with Information-Seeking Effectiveness (ISE), a local diagnostic defined as the fraction of help interactions followed by a correct next navigation step (not a final-task metric), rising from 50% to 74%. Three diagnostic contrasts fail to reproduce this structure. A separability test shows that the information-seeking pattern (mode split, ISE ranking) persists when answer quality is degraded (-18.8% accuracy), supporting an empirical separation between where an agent seeks help and the quality of the help it receives. Under the controlled answer channel, accuracy gains reach +16.2% at 10-digit; we read this as an upper bound on what better localization could unlock, not a deployment estimate.
arXiv:2606.08090v2 Announce Type: replace Abstract: Evaluating a natural-language yes/no predicate over a document corpus under an accuracy target - the semantic filter - is a cornerstone of LLM-based data processing. Calling the LLM on every document (the oracle) is prohibitive, so cascades pair the oracle with a fast proxy. As deployed today, they leave four limitations on the table. (1) Each cascade family - model-free clustering, prebuilt small-LLM proxies, online-trained proxies - commits to a single representation and pipeline, and wins on only a narrow query regime. (2) The strongest online proxy invests in a custom training scheme on a bi-encoder over dense embeddings, missing the token-level evidence richer predicates require. (3) The proxy is trained against binary yes/no labels, wasting the LLM's per-document confidence at the boundary documents it most needs to learn. (4) Existing calibrations add a uniform safety margin, conflating genuine proxy uncertainty with small-sample noise and inflating cascade cost. We address these by (1) composing families adaptively - model-free clustering first, online proxy only when needed, with oracle calls shared across phases; (2) replacing the cosine bi-encoder with a hybrid of off-the-shelf token-aware models; (3) training the proxy with the oracle's per-document confidence as a soft label; and (4) a calibration that adds the safety margin only where the labeled sample is sparse. We are also the first to use the oracle's per-document confidence for three purposes: a query-level difficulty compass, a lower bound on the minimum oracle calls any proxy-based cascade can make, and the proxy's soft training label. At a 90% accuracy target on three 10K-document corpora, our methods are 1.6-2.0x faster than the best prior method per corpus and meet the target on 95% of queries; the BER-derived lower bound indicates a further ~4-20x of headroom for future work.
arXiv:2606.08583v2 Announce Type: replace Abstract: Deep learning on physiological time series is interpreted through domain-specific features -- oscillatory rhythms in EEG, morphological complexes in ECG -- yet these signals sit atop a broadband aperiodic 1/f-like envelope that covaries with arousal, age, and pathology. We introduce a spectral audit framework combining aperiodic/periodic decomposition, phase-preserving Fourier interventions, sham controls, and simulation validation. Aperiodic reliance was task-dependent and architecture-general: across six neural architectures, flattening drops exceeded 0.42 balanced-accuracy points for sleep-wake classification, reached 0.07-0.13 for clinical abnormality detection, and remained minimal for motor imagery. Six of seven EEG foundation models showed FDR-significant aperiodic reliance on clinical EEG; age/sex and recording-era controls reduced but did not eliminate the effect. Applying the audit to PTB-XL ECG revealed neural drops of 0.32--0.36 persisting after demographic matching, confirming this confound class extends beyond EEG. Aperiodic controls should become standard for interpretable physiological time-series deep learning.
arXiv:2606.09150v2 Announce Type: replace Abstract: While recent autoregressive video diffusion models achieve remarkable streaming quality, they remain confined to low resolutions (e.g., 480P), leaving efficient, scalable, real-time high-resolution video generation a fundamental open challenge. To bridge this gap, we present Ultra Flash, a cascaded streaming framework capable of real-time high-resolution video generation. Ultra Flash achieves ~30 FPS at 1K resolution and ~18 FPS at 2K resolution on a single GPU through three key contributions: (1) an architecture-preserving T2V-to-TV2V super-resolution training paradigm coupled with an AIGC-oriented data degradation pipeline that effectively preserves the generative capability of the base model, enabling enhanced high-resolution detail when cascaded after mainstream low-resolution generative models; (2) a causal streaming latent upsampler paired with a high-resolution decoder, which enhances spatiotemporal coherence while enabling efficient latent spatial scaling and precise high-resolution decoding with negligible computational overhead; and (3) a cascade high-resolution streaming video generation optimization scheme that first performs hybrid-reward-enhanced sparse causalization and single-step distillation of the super-resolution model, then introduces cascaded streaming self-forcing preference optimization with dynamic cache management, jointly enhancing overall coherence, improving quality, and enabling real-time high-resolution streaming video generation. Extensive experiments demonstrate that Ultra Flash reliably produces ultra-high-resolution streaming video while maintaining state-of-the-art visual quality and superior efficiency. Project Page: https://xin1u.github.io/UltraFlash/
arXiv:2604.19586v2 Announce Type: replace-cross Abstract: Kirigami, the art of paper cutting, has been widely used in the modern design of mechanical metamaterials. In recent years, many kirigami-based metamaterials have been designed based on different planar tiling patterns and applied to different science and engineering problems. However, it is natural to ask whether one can create deployable kirigami structures based on the simplest forms of tilings, namely the monotile patterns. In this work, we answer this question by proving the existence of periodic and aperiodic monotile kirigami structures via explicit constructions. In particular, we present a comprehensive collection of periodic monotile kirigami structures covering all 17 wallpaper groups and aperiodic monotile kirigami structures covering various quasicrystal patterns as well as polykite tilings. We further perform theoretical and computational analyses of monotile kirigami patterns in terms of their shape and size changes under deployment. Altogether, our work paves a new way for the design and analysis of a wider range of shape-morphing metamaterials.
arXiv:2606.14061v2 Announce Type: replace Abstract: Coding agents powered by large language models have demonstrated strong performance on software engineering tasks. Yet most agents consume repositories almost entirely as text, which differs from how human developers use visual structure such as folder hierarchies and dependency relationships to orient themselves in large codebases. With multimodal large language models (MLLMs), it is an open question whether agents can effectively benefit from visual representations of repositories. This paper presents the first systematic empirical study of visual repository representations for LLM-based agents on repository-level issue resolution. We evaluate four recent multimodal models. Our results show that a strictly vision-only setup degrades accuracy and increases token cost, because agents lack sufficient symbolic detail and compensate with repeated visual queries. In contrast, integrating visual graphs of repository structure as a supplementary modality alongside standard text interfaces helps agents understand structure more efficiently: input token consumption decreases by up to 26% while issue-resolution accuracy is maintained or improved. Visualization is most useful during fault localization and when the agent autonomously controls exploration depth. These findings point to a practical hybrid text-and-vision design for next-generation coding agents.