arXiv:2607.07508v1 Announce Type: new Abstract: Reinforcement learning (RL) is becoming increasingly important for post-training large language models (LLMs). Previous RL pipelines for LLMs were mostly synchronous and batch-interleaved, which is inefficient for long-horizon agentic tasks. Recently, asynchronous RL has emerged as a more efficient alternative by updating the model as rollouts arrive. However, existing asynchronous RL systems often emphasize throughput, while leaving training stability and task effectiveness largely underexplored. For example, a key challenge is that group-wise sampling in the widely-used GRPO framework does not naturally fit asynchronous agentic training. In this paper, we present Single-rollout Asynchronous Optimization (SAO) to address the stability and off-policy challenges in asynchronous RL. To reduce off-policy effects and improve generalization, we replace group-wise sampling with single-rollout sampling, that is, using one rollout per prompt. We further improve this single-rollout strategy with practical value-model training designs. To improve optimization stability, we introduce a strict double-side token-level clipping strategy. SAO is able to train stably for one thousand steps and consistently outperform GRPO and its variants on agentic coding and reasoning benchmarks, such as SWE-Bench Verified, BeyondAIME, and IMOAnswerBench. We also demonstrate that single-rollout RL is particularly effective in a simulated online learning setting, where the model must adapt to changing evolving environments. To this end, SAO is successfully deployed in the agentic RL pipeline for training the open GLM-5.2 model (750B-A40B).
Science Journals
arXiv:2607.06740v1 Announce Type: new Abstract: Soft robots have attracted significant attention in applications such as medical intervention, rehabilitation, and robotic manipulation due to their inherent compliance, flexibility, and high degrees of freedom. Modular soft robots (MSRs), composed of multiple interconnected segments, represent an emerging class of robotic systems with highly deformable and reconfigurable structures capable of performing complex tasks. However, designing controllers for MSRs remains challenging due to their nonlinear dynamics, modeling complexity, and hyper-redundant nature. Existing approaches typically require controllers to be retrained from scratch whenever the robot morphology changes. In this work, we address these challenges through a continual learning inspired control framework capable of incrementally adapting to changes in robot morphology while preserving previously acquired knowledge. Specifically, the proposed framework enables the controller to sequentially learn new MSR configurations without forgetting previously learned ones. In addition, for MSRs with fixed configurations, the same framework can be employed in a distributed manner to learn module-specific dynamics, enabling localized control and improved precision. The proposed approach is validated through closed-loop trajectory tracking experiments in simulation using a tendon-driven soft robot, as well as on a real-world three-module pneumatic soft robotic arm. Furthermore, we demonstrate the adaptive capabilities of the framework through a reaching experiment in which the controller selectively activates only the necessary modules to reach a virtual target position, thereby reducing computational overhead.
arXiv:2607.07578v1 Announce Type: new Abstract: Chiral patterns have been observed in various processes from swirling bacterial colonies to tissue morphogenesis and cytoskeletal organization, yet the physical mechanisms underlying chiral cell motion remain poorly understood. Motivated by experiments demonstrating directional bias in the circular motion of confined cells, we use the tools of dynamical systems analysis with computer simulations to identify minimal intrinsic and extrinsic mechanisms capable of generating persistent biased migration. The dynamical systems framework reveals a common organizing principle: directional bias emerges through changes in the stability and/or basins of attraction of the clockwise and counter-clockwise motility states. We find four distinct routes to such bias. First, intrinsic torque in a polarized cytoskeleton can be spatially integrated to produce biased circular motion. Second, anisotropic cell-substrate friction can generate directional preference when reduced friction along the polarity axis is coupled to a directional offset. Third, a chiral wall-alignment response can also produce a persistent directional preference. Finally, substrate patterns that break mirror symmetry, such as dextral or sinistral ridges and troughs, can likewise bias rotational direction. Together, these mechanisms yield distinct, testable predictions and suggest a unifying lens for experimental interrogation of cellular chirality and the design of synthetic systems with programmable chiral motion.
arXiv:2604.28073v2 Announce Type: replace Abstract: Computer architecture simulation is essential for evaluating new designs without the need for costly tapeout. The community has developed dozens of valuable simulators that have enabled significant architectural advances. However, using and developing simulators remains a major barrier due to ad-hoc component interfaces, strict deployment requirements, the burden of managing performance optimizations like parallelization at the component level, and limited monitoring and visualization capabilities. The root cause of these limitations is the systematic neglect of user and developer experience in favor of technical functionality. We believe that only by separating technical concerns from user and developer experience concerns -- through a dedicated simulation engine decoupled from hardware models -- can the community overcome these fundamental obstacles and enable more productive architectural research. Akita embodies this philosophy as a dedicated simulation engine that cleanly separates infrastructure from architectural models. Smart Ticking and Availability Backpropagation let developers write simple cycle-based code while achieving event-driven performance. Parallel simulation happens transparently -- developers write single-threaded code while Akita handles multi-core execution. Akita's simple, uniform, yet powerful simulation tracing support enables real-time monitoring and post-simulation visualization. We demonstrate the flexibility of Akita through case studies, including the development of a trace-based DNN simulation and a RISC-V CPU simulation, showing how prioritizing developer experience accelerates architectural research.
arXiv:2605.03195v2 Announce Type: replace Abstract: Modern coding agents increasingly delegate specialized subtasks to subagents, which are smaller, focused agentic loops that handle narrow responsibilities like search, debugging or terminal execution. This architectural pattern keeps the main agent's context window clean by isolating verbose outputs (e.g. build logs, test results, etc.) within the subagent context. Typically when agents employ subagents for such tasks, they use frontier models as these subagents. In this paper, we investigate whether a finetuned small language model (SLM) can achieve comparable performance to frontier models in the task of agentic terminal execution. We present Terminus-4B, which is a post-trained Qwen3-4B model via Supervised Finetuning (SFT) and Reinforcement Learning (RL) using rubric-based LLM-as-judge reward, specifically for this task. In our extensive evaluation spanning various frontier models, training ablations and main agent configurations, we find that Terminus-4B is able to reduce the token usage of the main agent by up to ~30% compared to the No Subagent baseline with no impact to agent performance on benchmarks like SWE-Bench Pro and our internal SWE-Bench C# benchmark, which tends to be heavy in verbose execution tasks. Furthermore, Terminus-4B improves key metrics showing the main agent relying on the outputs of the subagent and doing fewer terminal execution tasks by itself. We see that our model not only closes the gap between the Vanilla Qwen model and frontier models like Claude Sonnet / Opus / GPT-5.3-Codex, but often even exceeds their performance.
arXiv:2605.19887v2 Announce Type: replace Abstract: Current Physical AI (PAI) relies heavily on closed-loop visual-servoing pipelines, whose perception and planning stages may become computationally intensive onboard due to complex models embedded on robots. In practice, offloading the perception task to on-site edges statically is inappropriate for latency-sensitive, precise industrial settings over a standardized industrial network. This emphasizes the importance of Control-Communication-Computing (3C) co-design in industrial automation: monolithic local execution saturates AI-accelerated machine and robot hardware, while static edge offloading exposes the control loop to network jitter. Existing adaptive task placement (ATP) controllers can partially address the gap by relocating a single pipeline stage on binary threshold rules, without a multi-stage model and an explicit cost on placement switching. In this paper, we propose a directed acyclic graph (DAG) based quality-of-service (QoS)-aware dynamic task placement (DTP) framework for sensing-perception-planning-control pipelines in networked robotics. This pipeline is formalized as a DAG with task-level and node-level attributes for compute cost, communication delay, and feasible placement sets; over a small interpretable candidate set (fully local, static offload, hybrid), a window-based cost function combines tail end-to-end latency, deadline violation rate, hardware utilization, and a Hamming-distance switching penalty, and a DTP algorithm with hysteresis and a minimum dwell-time bounds placement chatter. Our work presents the theoretical framework, a structured qualitative analysis, and a two-phase simulation plus hardware-in-the-loop validation roadmap.
arXiv:2607.07144v1 Announce Type: new Abstract: The key-value (KV) cache dominates the memory cost of long-context autoregressive inference, and a growing body of work compresses it through quantization, eviction, or offloading. We study a complementary question: once a position's KV state has been quantized to codebook indices, how should the resulting symbol stream be stored, and can the storage layer do more than store? A family of contractive iterated-map codes that serialize a symbol sequence into a sequence of low-dimensional real vectors is revisited, and it is shown that they form a natural archive format for a quantized KV cache with the following features. The method provides exactly the access pattern a growing cache requires. It is lossless, it runs in linear time, and supports O(1) random access and O(1) amortized append. A controlled study of the quantizer feeding this archive is conducted on GPT-2 with 1024-token contexts. Keeping a small exact window (4 attention sinks + 32 recent tokens) and archiving the rest, per-head residual vector quantization reduces the archived cache by 36-54x relative to an fp16 cache at a perplexity cost of 11-15%, and we quantify a sharp key/value asymmetry -- quantizing keys is roughly 4x more damaging than quantizing values, consistent with prior low-bit KV work -- and use it to allocate bits in a hybrid scheme. Finally, we show the archive is simultaneously a search index: approximate substring queries execute directly on the stored vectors, and matched context is decoded from the matched vector without ever materializing the surrounding text. We release all code; every number reproduces from a single command on a laptop CPU.
arXiv:2607.07580v1 Announce Type: new Abstract: While echocardiography is essential for cardiovascular diagnosis, inherent speckle noise and low signal-to-noise ratio often lead to ambiguous semantic features and fragmented boundaries. These limitations significantly hinder the segmentation accuracy of deep learning models in complex clinical cases. Moreover, temporal motion of the heart plays a critical role in recognizing anatomical structures. To address these challenges, we designed a STLSF module which comprises a window-matching-based semantic correction component and a semantics-guided texture enhancement component. By leveraging local transition probability correlations to correct semantics and employing semantics-guided texture enhancement, the STLSF module effectively mitigates texture instability and ambiguous semantic interpretations caused by disadvantaged echocardiography quality. Additionally, to facilitate the encoder's adaptation to the intrinsic priors of ultrasound-specific imaging patterns, we propose a frequency-aware denoising pre-training method. The entire work builds a convolution-based network with locality inductive bias and long-range dependencies. Extensive experiments confirm our SOTA performance, achieving 93.87\% Dice on CAMUS and 92.62\% on EchoNet-Dynamic, with respective HD95 values of 3.29mm and 2.73mm.
arXiv:2607.07330v1 Announce Type: new Abstract: Hypergraph neural networks have shown powerful capability in modeling higher-order relations, yet their predictive uncertainty remains underexplored. Unlike pairwise graphs, uncertainty in hypergraphs arises not only from noisy attributes and ambiguous labels, but also from variations in node-hyperedge incidence structures and complex higher-order dependencies. Existing approaches mainly estimate uncertainty from final predictions or rely on computationally expensive ensembles and Bayesian inference, limiting their ability to capture uncertainty evolution during representation learning. In this paper, we propose Hypergraph Neural Stochastic Diffusion(HyperNSD), a stochastic differential equation framework for uncertainty estimation on hypergraphs. HyperNSD models hypergraph representations as stochastic processes evolving over node-hyperedge incidence structures. A learnable drift function captures deterministic higher-order diffusion dynamics, while a learnable stochastic forcing function characterizes structural ambiguity and representation noise. Predictive uncertainty is directly quantified through the variability of stochastic representation trajectories, providing an intrinsic uncertainty measure beyond post-hoc confidence scores. We formulate HyperNSD with neural drift and diffusion networks, enabling joint learning of prediction and uncertainty propagation. Theoretical analyses establish well posedness, perturbation stability,permutation equivariance, and numerical convergence of the proposed stochastic dynamics. Experiments on multiple hypergraph benchmarks demonstrate that HyperNSD achieves reliable uncertainty estimation for out-of-distribution and misclassification detection while preserving competitive prediction accuracy. These results provide a principled stochastic-dynamical framework for trustworthy higher-order representation learning.
arXiv:2607.07502v1 Announce Type: cross Abstract: Isospectrally patterned lattices exhibit a composite band structure with a tunable ratio of localized versus delocalized eigenstates that is controlled by the underlying phase gradient. We show that the lattice Hamiltonian can be interpreted as that of a single spin exposed to a rotating magnetic field which is allowed to hop with a spin-flip across the lattice. In the low- and high-energy part of the band the localized states show an envelope of oscillatory character separated by quasi-nodes. Spin peaks occur at the locations of these quasi-nodes and provide a unique spin texture to the eigenstates which becomes increasingly complex with increasing degree of excitation. The crossover from localization to delocalization and vice versa leaves its fingerprints in the Fourier spectrum of the eigenstates: the original bimodal frequency distribution widens with increasing degree of excitation, moves across the spectral window and finally culminates in an extremely narrow frequency peak. In the course of this evolution the spin texture undergoes a rearrangement transition involving different characteristic (ir)regular patterns which we quantify by considering the total variation of the local spin fluctuations. Our results demonstrate the variety of the spectral properties of isospectrally patterned lattices which holds great prospect in particular when considering higher lattice or cell dimensions.
arXiv:2607.07419v1 Announce Type: new Abstract: Biplex, as a relaxation of the biclique model, has emerged as an important cohesive subgraph model for bipartite graph analysis. The maximum $k$-biplex search problem aims to identify the $k$-biplex with maximum number of edges and has been widely applied in various real-world applications, including community detection, online recommendation, and fraud detection. However, the problem is NP-hard, and existing exact algorithms remain inefficient on large-scale bipartite graphs with large values of $k$ (e.g., $k\geq 3$). In this paper, we revisit the maximum $k$-biplex search problem from a complementary perspective. We reveal a novel structural duality: finding a maximum $k$-biplex in a bipartite graph is equivalent to finding a minimal $k$-bounded-degree deletion in its complement graph. Based on this observation, we propose a novel deletion-based algorithm for the maximum $k$-biplex search problem. We theoretically prove that the proposed algorithm achieves a worst-case time complexity of $O^*(\gamma_k^n)$, where $\gamma_k<2$. Specifically, $\gamma_1=1.725$, $\gamma_2=1.856$, and $\gamma_3=1.928$. To further enhance practical efficiency, we develop several effective upper-bounding techniques and a heuristic strategy for obtaining high-quality initial solutions, which substantially reduce the search space. Extensive experiments on eight real-world bipartite graphs demonstrate the efficiency of our approach, which achieves up to four orders of magnitude speedups over state-of-the-art algorithms.
arXiv:2601.00357v2 Announce Type: replace Abstract: As adversaries increasingly weaponize encryption and protocol obfuscation to evade traffic detection, traditional methods are rendered obsolete, necessitating deep learning to unmask sophisticated threats. However, the prohibitive computational costs of existing large models create a critical defense gap, hindering their deployment in real-time and throughput-sensitive environments. To close this vulnerability, we introduce Traffic-MoE, a sparse foundation model tailored for traffic security analysis. By dynamically routing traffic tokens to a small subset of specialized experts, Traffic-MoE effectively decouples model capacity from computational overhead. Extensive evaluations across four security-oriented tasks demonstrate that Traffic-MoE achieves state-of-the-art or highly competitive performance compared to leading competitors. Crucially, it delivers a 70.42% increase in throughput, reduces inference latency by 41.39% while significantly optimizing GPU memory consumption. Beyond efficiency, Traffic-MoE exhibits superior robustness against adversarial traffic shaping and maintains strong detection capabilities in few-shot scenarios, establishing a scalable and resilient paradigm for modern network traffic security analysis.
arXiv:2601.06135v3 Announce Type: replace Abstract: Spatial computation in geographic systems increasingly requires query-conditioned, local, interpretable aggregation under metric constraints. Many classical approaches rely on global summation and treat approximation as an implementation concern, limiting interpretability and scalability at large scales. We propose the Adaptive Density Field (ADF), a geometric attention framework that formulates spatial aggregation as a query-conditioned, metric-induced attention operator in continuous space. Given a set of labelled spatial points with associated scalar scores, ADF defines a continuous intensity field over space. For a given query location, the field value is obtained via a local adaptive Gaussian kernel mixture centered on the query's nearest neighbors, where kernel bandwidths are modulated by point-specific scores to evaluate local aggregated influence. Additionally, approximate nearest-neighbor search is introduced, enabling scalable execution while preserving locality. The proposed ADF bridges concepts from adaptive kernel methods, classical GIS methods, and attention mechanisms by reinterpreting spatial influence as geometry-embedded attention, grounded in physical distance rather than learned latent projections. The proposed framework is formulation-level rather than algorithm-specific, allowing flexible kernel choices, score-to-bandwidth mappings, and approximation parameters. This approach provides a unifying perspective on spatial influence modeling that emphasizes structure, scalability, and geometric interpretability, with relevance to geographic information systems and spatial machine learning.
arXiv:2502.09978v4 Announce Type: replace Abstract: Internet of Things (IoTs) have been widely applied in Collaborative Intelligent Transportation Systems (C-ITS) for the prevention of road accidents. As one of the primary causes of road accidents in C-ITS, the efficient detection and early alarm of road hazards are of paramount importance. Given the importance, extensive research has explored this topic and obtained favorable results. However, most existing solutions only explore single-modality data, struggle with high computation and communication overhead, or suffer from the curse of high dimensionality in their privacy-preserving methodologies. To overcome these obstacles, in this paper, we introduce RoadFed, an innovative and private multimodal Federated learning-based system tailored for intelligent Road hazard detection and alarm. This framework encompasses an innovative Multimodal Road Hazard Detector, a communication-efficient federated learning approach, and a customized low-error-rate local differential privacy method crafted for high dimensional multimodal data. Experimental results reveal that the proposed RoadFed surpasses most existing systems in the self-gathered real-world and CrisisMMD public datasets. In particular, RoadFed achieves an accuracy of 96.42% with a mere 0.0351 seconds of latency and its communication cost is up to 1,000 times lower than existing systems in this field. It facilitates collaborative training with non-iid high dimensional multimodal real-world data across various data modalities on multiple edges while ensuring privacy preservation for road users.
arXiv:2503.15581v2 Announce Type: replace Abstract: Real-time safety assessment is critical for ensuring the reliable operation of complex dynamic systems. However, obtaining full safety labels in real time is often prohibitively expensive, resulting in a challenging mixed-feedback scenario dominated by partial feedback, especially under concept drift. Furthermore, existing online ensemble methods typically rely on heuristic weight allocation, lacking provable performance guarantees under such limited-feedback conditions. To address these challenges, we propose PB-OEL, a performance-bounded online ensemble learning framework designed for real-time safety assessment under mixed feedback. At the ensemble level, a theoretical framework is established to bound the performance of the ensemble classifier relative to its base classifiers across varying feedback ratios. By formally defining the form of expert advice, the bound guarantees that the ensemble outperforms any individual base classifier over a sufficiently large data stream. At the base-classifier level, a penalty-based update strategy is introduced, enabling base models to explicitly leverage misclassified samples rather than simply discarding them. Extensive experiments on the real-world Jiaolong manned submersible dataset demonstrate that PB-OEL maintains robust predictive performance and outperforms state-of-the-art methods.
arXiv:2606.20374v2 Announce Type: replace Abstract: Large-scale LLM training requires always-on, fine-grained observability for effective performance diagnosis at scale. Coarse resource monitors alone cannot localize root causes, and fine-grained profilers incur prohibitive (5%-30%) overheads and massive trace volumes, making always-on deployment impractical in large production clusters. We propose ARGUS, a low-overhead, fine-grained, always-on tracing and real-time analysis system for training workloads in 10,000+ GPU-scale production clusters. ARGUS decomposes observation along the training call hierarchy into CPU call stacks, framework semantics, and GPU kernel execution, with always-on collection under a combined overhead of less than 2%. It builds a unified data pipeline and compresses raw kernel events by approximately 3,700x from 10 MB to 2.7 KB per rank per step. Its progressive diagnosis framework automatically isolates anomalous windows, straggler ranks, and degraded kernels through iteration-time, phase-level, and kernel-level analysis. Deployed for over six months on a 10,000+ GPU production cluster, ARGUS has supported continuous fail-slow detection and performance optimization. Our case studies further demonstrate its effectiveness across representative anomalies, including compute stragglers, link degradation, pipeline-bubble amplification, FlashAttention JIT stalls, and compute stragglers masked by communication symptoms.
arXiv:2607.05419v2 Announce Type: replace Abstract: Accurate and timely channel state information (CSI) is essential for next-generation wireless systems, yet existing works treat CSI compression and CSI prediction as separate problems, both in academia and in current 3GPP studies. Consequently, channel aging remains insufficiently addressed within standardized CSI feedback pipelines. In this article, we propose a unified compression-prediction framework that integrates Contrastive Predictive Coding (CPC) directly into the 3GPP-compliant CSI compression architecture. Instead of predicting high-dimensional CSI matrices, our approach forecasts future latent representations and jointly optimizes reconstruction fidelity and temporal predictive coherence via a combined 1-SGCS and InfoNCE objective. This design enables temporal representation learning without increasing feedback overhead. We present two variants: CPC-before-Compression, which performs autoregressive modeling on encoded features prior to quantization, and CPC-after-Compression, which shifts temporal modeling to the base-station to reduce the complexity of users' devices. Evaluations on 3GPP-compliant datasets from Nokia, Oppo, and CATT show that CPC-before-Compression achieves over 90% reconstruction accuracy with 32x lower decoder GFLOPs than the 3GPP baseline, while CPC-after-Compression preserves an identical encoder footprint and the same 64-bit feedback overhead. By unifying compression and prediction within a standardized pipeline, the proposed framework provides an age-aware, computationally efficient CSI feedback solution. The source code is publicly available at: https://github.com/AhmedRadwan02/cpc-3gpp
arXiv:2605.09778v2 Announce Type: replace Abstract: Evaluating softmax attention over a fixed long context requires reading every cached key-value pair for each new query token. For a given context (a book, a manual, a legal corpus) the attention output is a deterministic function of the query. We propose Nectar, which fits a compact neural network to this function for queries drawn from a task-relevant distribution. Nectar fits two networks per layer and KV-head: a target network that predicts the attention output and a score network that predicts the log-normalizer. The pair plugs into the standard masked self-attention at inference time, replacing the $O(n)$ attention over the cache with a forward pass whose cost does not depend on $n$. Each module carries on the order of $|\theta|$ parameters per layer and KV-head, typically much smaller than the $2nd$ KV-cache footprint at the same granularity. We report experiments on models from 1.7B to 8B parameters across five long-context datasets. The approximation error tracks the next-token accuracy gap to full attention, and allocating capacity non-uniformly across layers reduces that gap in our ablation. Beyond this analysis of metrics, we check that the text generations (following a question prompt) of a model equipped with a Nectar module match in semantic content those obtained by giving the same model access to the full cache.
arXiv:2605.10240v3 Announce Type: replace Abstract: Software vulnerability detection is critical for ensuring software security and reliability. Despite recent advances in deep learning, real-world vulnerability datasets suffer from two severe challenges: frequency imbalance and difficulty imbalance. We reinterpret these challenges from an embedding geometry perspective, observing that such imbalances induce geometric distortions in hyperspherical representation space. To address this issue, we propose MARGIN, a metric-based framework that learns discriminative vulnerability representations through adaptive margin metric learning and hyperspherical prototype modeling. MARGIN dynamically adjusts geometric regularization according to the distribution structure estimated by the von Mises-Fisher concentration, aligning the probability mass of embedding distributions with their corresponding Voronoi cells, thereby reducing geometric distortion and yielding more stable decision boundaries. Extensive experiments on public vulnerability datasets show that MARGIN consistently outperforms strong baselines, achieving notable improvements in classification and detection, especially on challenging, imbalanced datasets. Further analysis demonstrates that MARGIN produces more structured embedding geometries, improving robustness, interpretability, and generalization.
arXiv:2607.06447v2 Announce Type: replace Abstract: Recent LLM-based mathematical reasoning agents have begun to tackle research-level problems and, in several cases, have contributed to the resolution of open problems. However, scaling and orchestrating such agents effectively remains challenging, due to the difficulty of coordinating parallel proof search while keeping intermediate claims organized and reliable. In this paper, we propose Danus, an orchestration system for research-level mathematical reasoning centered on a shared fact graph as a global memory-management mechanism. Danus consists of a main agent that performs planning and coordination, multiple worker agents that carry out proof search in parallel, and a stateless verifier that checks proposed mathematical claims before they are admitted into the fact graph. Each verified fact is stored together with its proof and logical dependencies, allowing the system to build long arguments incrementally while keeping the shared proof state organized. The main agent periodically summarizes the evolving proof state, redirects workers across promising directions, and supports interaction with human mathematicians through progress reports. We evaluate Danus through six research-level case studies in algebraic geometry, singularity theory, and combinatorics, illustrating how the fact-graph memory mechanism enables Danus to construct long, detailed mathematical proofs. Our results suggest that fact-graph-based orchestration provides an effective route toward scaling mathematical reasoning agents for long-horizon research problems. Danus is open source at https://github.com/frenzymath/Danus.
arXiv:2605.17842v3 Announce Type: replace Abstract: Autoregressive language models execute Transformer layers sequentially, creating a latency bottleneck that is not removed by conventional tensor or pipeline parallelism. We study whether this layerwise dependency can be relaxed by treating the hidden-state trace across layers as the solution of a nonlinear residual equation and solving it with parallel Newton-style updates. While this view is principled, exact Newton corrections require expensive Jacobian-vector products and naive fixed-point iterations are unstable on trained Transformers. We introduce Structured Newton Layer Parallelism (SNLP), a training and inference framework that replaces exact layer Jacobians with cheap architecture-induced surrogate dynamics. In residual Transformers, this yields Identity Newton (IDN), where the correction reduces to a prefix-sum-like update; in mHC-style architectures, HC Newton (HCN) uses the model's residual mixing matrix. We also study SNLP-aware training, including pretraining regularization and direct SNLP-forward SFT. Experiments on Nanochat-scale Transformers show that SNLP exposes a practical speed-quality frontier: on 0.5B models, it reaches up to 2.58x wall-clock speedup, and a less aggressive configuration reaches 1.40x speedup without increasing PPL. The useful tradeoff comes from the biased finite-iteration computation induced by IDN/HCN rather than exact recovery of the sequential trace. We further show that SNLP-forward SFT can preserve downstream task accuracy, and that SNLP can serve as a drafter for self-speculative decoding while a sequential verifier preserves output correctness.
arXiv:2607.07025v1 Announce Type: new Abstract: In this paper, we derive a new model to simulate the incompressible resistive magnetohydrodynamic (MHD) free surface flow. A thermodynamically consistent diffuse interface method is adopted to characterize the moving interface in the modeling process. The formal convergence of the proposed MHD free surface flow model to the sharp interface model is established via a matched asymptotic argument, and the model can be solved without the need for sophisticated free surface capturing schemes. We design a fully decoupled linear finite element scheme that preserves the divergence-free constraint of the magnetic field at a discrete level. The reliability and robustness of the proposed model and algorithm are validated through numerical investigations of the magnetic damping effect on bubble dynamics. In particular, we provide a quantitative numerical comparison of the present results with those obtained from an inductionless MHD model and a sharp interface arbitrary Lagrangian--Eulerian model.
arXiv:2606.03565v4 Announce Type: replace Abstract: LLM agents often solve complex tasks by composing skills, making skill retrieval a front-end component of agent systems. Unlike document retrieval, top-K correctness in skill retrieval depends not only on the relevance of each query-skill pair, but also on whether the retrieved skills can work together under the query. This query-conditioned "skill compatibility" cannot be recovered from independent relevance alone. However, LLM-based synthesis pipelines already produce a useful signal for it: the LLM's own rejection decisions, which specify which skills should not be retrieved together for a given query, but are usually discarded as low-quality data. We propose Reject-as-Resource Retriever (R3) and construct R3-Skill, a bilingual (Chinese-English) benchmark for agent skill routing. R3-Skill covers four language directions and uses LLM-rewritten queries that better approximate user requests; its test-set ground truth is verified by multiple experts. It contains 10,246 skills grouped into 8 thematic super-domains, 41,592 accepted queries, and 32,828 LLM-rejected annotations, further organized into an 8-class rejection-reason taxonomy. R3-Skill keeps this normally discarded rejection signal and uses it as compatibility supervision. On R3-Skill, we train a two-stage retriever consisting of R3-Embedding and R3-Reranker. Gradient analysis explains why this query-conditional signal is weak when injected into the tested bi-encoder objective under bilateral balancing, while a cross-encoder can use it as graded ranking supervision; R3-Skill ablations support this split. The R3-Embedding + R3-Reranker pipeline reaches Hit@1 = 0.7521, NDCG@10 = 0.8173 and Set-Compat = 0.3188 on R3-Skill. The dataset, model weights, and evaluation scripts will be open-sourced.
arXiv:2607.07073v1 Announce Type: new Abstract: Searching for time-series segments that match user-defined patterns is important in domains such as finance, climate science, and healthcare. However, existing visual query tools often struggle to support vague, composite, or fuzzy pattern descriptions, often requiring users to express their intent through precise sketches or rigid structured filters. We present ShapeTalk, a coordinated natural-language and sketch-based querying system for univariate time-series pattern search. Rather than treating text and sketch as a fused input stream, ShapeTalk uses them as complementary representations of analytic intent: natural language supports semantic and compositional pattern descriptions, while sketching supports direct geometric refinement. The two modalities are linked through a shared visual context, editable feature representations, and synchronized result views, enabling users to move between text and sketch during iterative query formulation. At its core is an LLM-based semantic parsing pipeline that translates free-form natural-language queries into interpretable and editable shape-feature constraints. We evaluate ShapeTalk through two usage scenarios, a user study with failure-case analysis, and an assessment of the LLM-based semantic parsing pipeline. The results show that ShapeTalk supports effective time-series pattern search, with natural language serving as an accessible entry point and sketching providing a complementary mechanism for refinement and recovery when textual specifications are insufficient.
arXiv:2607.07075v1 Announce Type: new Abstract: Entropy-based methods have long been used for network anomaly detection, but most existing approaches treat entropy as a scalar statistic on narrow observables rather than as part of a broader behavioral state-space for cyber systems. We propose a finite-dimensional macrostate framework for network telemetry, instantiated over the Canonical Security Telemetry Substrate (CSTS), so that coarse-graining is performed over persistent entities, typed relations, and temporal state rather than isolated event records. The resulting macrostate captures activity, distributional disorder, structural organization, temporal volatility, persistence, and deviation from benign baselines. Rather than scoring only unusual states, we model window-to-window macrostate transitions and define regime structure, stability, and anomalous change. This supports discrimination between benign workload drift and adversarial reorganization. We evaluate the framework on benchmark network telemetry datasets and compare it against Shannon-, R\'enyi-, and Tsallis-style entropy baselines, as well as standard anomaly detectors. The proposed representation improves anomaly discrimination and supports more interpretable behavioral analysis of cyber telemetry.