arXiv:2606.16814v1 Announce Type: new Abstract: Train rescheduling repairs disturbed timetables while enforcing train-path precedence, resource capacity, and delay objectives. Dynamic Discretization Discovery (DDD) avoids full time discretization by refining only time points needed to certify feasibility and optimality. We strengthen a recent MaxSAT-DDD model through two encoding changes. First, resource conflicts are encoded as time-dependent at-most-one cliques, using pairwise clauses for small cliques and a sequential counter for large cliques. Second, earliest feasible times are propagated along train paths before the first DDD iteration. We evaluate four MaxSAT variants, two SAT optimization backends, Gurobi/CPLEX MILP models, and CPLEX CP on 72 instances and three delay objectives. MaxSAT-DDD solves all stepwise instances in about 23 ms on average. MaxSAT-Default reduces rounded-cost runtime from 794 to 479 ms, and the ablation study reports up to 79.6\% runtime reduction on the common-solved subset of hard continuous track instances.
Science Journals
arXiv:2606.16817v1 Announce Type: new Abstract: Recent retrieval-augmented generation (RAG) approaches have demonstrated strong capability in handling complex queries, yet current research overlooks a critical challenge: different retrievers require fundamentally different query formulation strategies for optimal performance. In this work, we present the first systematic analysis of how LLMs can learn to adapt their query formulation strategies for different retrievers via reinforcement learning (RL). Our empirical study reveals that RL effectively teaches an LLM to tailor its queries to specific retriever characteristics. We discover that different retrievers exhibit surprisingly distinct optimal query styles (e.g., descriptive vs. question-like), suggesting strategies learned for one retriever ineffective for another. We further show that performance can be enhanced by incorporating retriever-specific human guidance and by scaling model size. To facilitate learning over multi-retrieval-step trajectories, we introduce a branching-based rollout technique that improves training stability. Our work provides the first empirical evidence and actionable insights for building truly retriever-aware RAG systems. Code and resources are available at https://github.com/LCO-Embedding/Envs-aware-Information-Retrieval.
arXiv:2606.16822v1 Announce Type: new Abstract: AI has become a partner in how people learn about, do, and engage with science, and the partnership takes three forms: a scientist works with a co-scientist whose output must be checked; a member of the public looks something up to decide whether a diet works or whether to fit solar panels; and a student takes up an inquiry with AI in a science class. Across all three, one thing decides whether the partnership helps or harms: whether the human evaluates what the AI returns or takes it on trust. I argue that this evaluation -- epistemic vigilance calibrated to how far a fallible source can be trusted -- is, given adequate prior knowledge, the binding constraint on productive augmentation. You can hand the AI a great deal precisely because you stay vigilant; vigilance makes generative partnership safe, so it licenses augmentation rather than restricting it. Vigilance is already invoked in science education but under-specified for the AI case; I specify its components, the mechanism tying it to learning, and a way to measure it without soliciting the evaluation it is meant to detect. What is distinctive is that the machine's fluent, confident prose reads as trustworthy whether or not it is, so its surface works against the human evaluating it. The argument bears hardest on education: the integrated conceptual knowledge instruction aims to foster forms only under deep processing, and vigilance sets how deeply a claim is processed, so it is the precondition for learning with AI. The design factors the field reports matter through whether they engage the learner's evaluation; none works around it. Untested is vigilance as a measured disposition, above all where the AI is confidently wrong. Because it is unevenly distributed, integrating AI uniformly is likely to widen the gap between better- and less-prepared students. I close on how it might be built by fading support as the learner takes over.
arXiv:2606.16824v1 Announce Type: new Abstract: Coding agents are a fast-growing LLM application, executing as long-running closed-loop sessions in which LLM generations alternate with external tool calls. Yet, unlike chat workloads, their serving behavior has not been studied extensively. We address this gap by collecting a dataset of real-world coding assistant traces. Our analysis shows that coding agent sessions repeatedly reuse large prefixes and create sustained KVCache pressure that conventional LLM serving policies handle poorly. Based on our analysis, we present CacheWise, a KVCache management layer that improves KVCache reuse for coding agent workloads. CacheWise combines prefix-aware scheduling with reuse-aware eviction guided by lightweight predictions from tool call metadata. Implemented in vLLM and evaluated on the collected traces, CacheWise reduces KVCache evictions by up to 2-2.6x and improves total agent session completion time by up to 3.5x.
arXiv:2606.16826v1 Announce Type: new Abstract: Generalist manipulation policies are increasingly presented as foundation models for robotic control, but their real-world generalization remains difficult to diagnose. A policy may succeed on demonstrated tasks while still failing to execute fine-grained atomic skills or recombine learned skills in new task structures. We introduce \textbf{ATOM-Bench}, a real-world benchmark for evaluating both atomic skills and compositional generalization in manipulation policies. ATOM-Bench factorizes tabletop manipulation into motor atoms and instruction atoms, and contains 30 atomic tasks and 24 held-out compositional tasks across paired single-arm and dual-arm robot tracks. We collect 3,000 human demonstrations for atomic fine-tuning and release both the demonstration data and evaluation rollout data to support reproducible real-world evaluation. Policies are fine-tuned on atomic tasks and evaluated on both atomic skill acquisition and held-out compositional tasks. We further introduce Atomic Score (AS) and Compositional Failure Share (CFS) to distinguish failures caused by weak atomic skills from failures caused by limited compositional reuse. Through 2,700 physical rollouts on five representative manipulation policies, we find that current policies can acquire simple instruction-grounding skills, but still struggle with fine-grained motor atoms, counting, and logical filtering. More importantly, strong atomic performance does not reliably transfer to held-out compositional tasks. ATOM-Bench provides a diagnostic testbed for studying whether failures arise from weak motor execution, poor instruction grounding, or limited compositional reuse.
arXiv:2606.16827v1 Announce Type: new Abstract: Large Language Models (LLMs) have significantly advanced the automation of software engineering tasks. One prominent example is code generation, where an LLM produces code in a specified programming language based on a natural language description. Most research in this area has focused on high-resource languages, such as Python or Java, which benefit from abundant training data. A smaller body of work has explored low-resource languages, which are underrepresented in training corpora. In contrast, no-resource languages for which LLMs have seen virtually no training data remain largely unstudied. These languages often emerge in industry, where organizations develop proprietary or domain-specific languages unsupported by commercial tools like GitHub Copilot. This results in the need for companies to deploy their own in-house code recommenders. To investigate possible solutions in this context, we build and release three code generation benchmarks for no-resource languages, based on two recently proposed programming languages for which very little training data is available. Using these benchmarks, we experiment several solutions to teach LLMs about no-resource languages, including prompt-based techniques as well as pre-training and fine-tuning exploiting the little data available. While further pre-training gives the largest performance gains for no-resource languages, applying it directly to instruction-tuned models harms their ability to follow instructions. To address this, we start from a base model, further pre-training it on the target language, and then inject instruction-following capabilities via weight diff transfer from an instruction model. Such an approach significantly improves code generation capabilities in no-resource settings, allowing companies to cheaply deploy a specialized instruct model without dealing with the computational cost of instruction fine-tuning.
arXiv:2606.16831v1 Announce Type: new Abstract: Block Markov superposition transmission (BMST) codes provide a flexible framework for constructing codes with near-capacity performance and low-complexity sliding-window decoding. However, existing BMST variants show contrasting performance limitations: recursive BMST (rBMST) codes suffer from error propagation but avoid high error floors, whereas non-recursive BMST codes exhibit the opposite behavior. Motivated by these complementary characteristics, we combine recursive and non-recursive components through parallel and serial concatenation, yielding mixed BMST (mBMST) codes. The proposed framework subsumes existing BMST variants and enables new BMST structures. Simulations show that these structures improve FER and BER performance with lower memory requirements than rBMST.
arXiv:2606.16836v1 Announce Type: new Abstract: Transformer Grammars (TGs) enhance language modeling by incorporating syntactic tree structures. Despite the potentially significant impact on model performance of how syntactic trees are linearized in TGs, existing studies rely solely on Depth-First Traversal (DFT) for linearization. In this paper, we expand the traversal design space by exploring Breadth-First Traversal (BFT) and a novel hybrid traversal strategy, Production-Rule Traversal (PRT), which combines the structural lookahead of BFT with the early lexical generation of DFT. We integrate these traversal methods with varying tree configurations and masking strategies, and empirically evaluate their performance on language modeling, syntactic generalization and summarization. We reveal the inherent trade-offs between nested composition and global lookahead, providing actionable recommendations for designing task-aware Transformer Grammars.
arXiv:2606.16837v1 Announce Type: new Abstract: Spoofed speech detection is increasingly challenged by realistic synthesis, voice conversion, and replay attacks, with cross-dataset generalization remaining a major limitation. This work we propose a Temporal Pyramid Adapter that utilize parallel temporal convolutions with varying receptive fields to capture multi-scale spoofing cues, ranging from local artifacts to global prosodic irregularities. We also integrated self-supervised XLS-R representations combined with front-end adapters, including Mel, Sinc, and a Temporal Pyramid design for multi-scale temporal modeling. The proposed model is evaluated cross multiple benchmark including ASVspoof 2017, ASVspoof 2021 (DF/LA), PartialSpoof, DiffSSD, and multilingual HQ-MPSD datasets. Experimental results demonstrate that Temporal Pyramid model obtained AUC of 99.24% and a EER of 3.87% on the PartialSpoof database, which is significantly outperforming the base model and several SOTA baseline such as LCNN-BLSTM (9.87% EER) and TRACE (8.08% EER). Additionally, multilingual evaluations confirm that while spoofing artifact are independent from language. While self-supervised representations improve robustness, performance degrades under domain and language shifts, highlighting the need for better adaptation and calibration strategies.
arXiv:2606.16838v1 Announce Type: new Abstract: Multi-task learning (MTL) is essential in recommender systems to enable complementary learning among diverse user feedback. While modern industrial practices have shifted from DNNs to Transformer-centric architectures to strengthen sequence modeling and scaling capacity, they still decouple feature encoding from multi-task prediction, treating the Transformer as a task-agnostic encoder. This design fundamentally limits the performance and scalability by (1) creating an information bottleneck under heterogeneous task objectives, (2) inducing gradient interference that leads to the seesaw phenomenon, and (3) forcing a dataflow transition in which attention-based, context-adaptive representation learning is converted to static feed-forward task prediction with incompatible information read-write dynamics. We propose OneRank, a Transformer-native multi-task ranking framework that eliminates encoder-predictor separation and introduces task-private channels for forward representation learning and backward optimization, enabling task-specialized learning while reducing inter-task interference. In the forward pass, OneRank learns task-specific representations bottom-up through task-conditioned information selection, candidate-aware contextualization, and controlled cross-task interaction. In the backward pass, cross-task gradient detachment isolates task-private parameter updates from shared knowledge extraction modules, preventing negative transfer. We further replace static task-specific MLP scorers with dynamic matching-based scoring for context-aware personalized ranking. By internalizing multi-task reasoning within the Transformer stack, OneRank establishes a unified and scalable architectural paradigm. Offline and online experiments on large-scale industrial datasets show that OneRank significantly outperforms state-of-the-art baselines while maintaining computational efficiency.
arXiv:2606.16841v1 Announce Type: new Abstract: The amplitudes of light modes in a homogeneous interface layer are investigated around the critical conditions (CC) in a total reflection geometry. CC occur when the normal wave vector vanishes; the resulting divergence upon variation of the angle of incidence is characterized by a critical exponent -0.5. Absorption replaces the divergence with a finite peak whose width and height are derived analytically. The high relevance of the amplification for Evanescent Wave Dynamic Light Scattering (EWDLS) is demonstrated using published experimental data. The relation to surface plasmon resonance (SPR) is briefly discussed. An outlook connects the amplitude divergence to a critical analysis of the Distorted Wave Born Approximation (DWBA) presented in a companion paper.
arXiv:2606.16850v1 Announce Type: new Abstract: This work presents a novel digital image correlation (DIC) framework for full-field measurements of displacement, strain, and damage, based on a phase field (PF) approach. The idea is to take advantage of the ability of the PF method to track complex crack morphologies and to provide a natural way in DIC to perform damage and crack measurements from experimental speckle images, in addition to displacement and strain fields. Moreover, incorporating the damage variable into DIC can improve the displacement accuracy near the crack tip, and can avoid the need of user-defined masks when dealing with cracked samples, which is advantageous when cracks become complex and the manual application of masks becomes challenging. The theoretical formulation of the proposed framework, namely PF-DIC, was presented in detail in the paper, along with a finite element implementation. Numerical examples have demonstrated the capability of the proposed PF-DIC in terms of capturing different types of cracks while improving the measurement accuracy of displacement fields. Additionally, it is shown that the PF-DIC can be easily adapted to selectively identify critical cracks under specific loading conditions or mechanisms for damage assessment and diagnostic purposes. The proposed DIC framework offers a powerful and automatic damage measurement technique and paves the way for the unification of PF simulations and experimental measurements of fracture and opens numerous opportunities in science and engineering, e.g., materials defect characterization and structural health monitoring.
Video-Based Optimal Transport for Feedback-Efficient Offline Preference-Based Reinforcement Learning
arXiv:2606.16856v1 Announce Type: new Abstract: Conveying complex objectives to reinforcement learning (RL) agents often requires meticulous reward engineering. Preference-based RL (PbRL) offers a promising alternative by learning reward functions from human feedback, but its scalability is hindered by high labeling costs. Inspired by advances in Video Foundation Models (ViFMs), we present Video-based Optimal Transport Preference (VOTP), a semi-supervised framework that learns effective reward functions from only a handful of labels. By leveraging optimal transport to align visual trajectories within the rich representation space of ViFMs, VOTP effectively generates high-fidelity pseudo-labels for large amounts of unlabeled data, substantially reducing human supervision. Extensive experiments across locomotion and manipulation benchmarks demonstrate the superiority of VOTP, which outperforms state-of-the-art offline PbRL methods under limited feedback budgets. We also showcase the robustness of VOTP in the presence of visual distractors and validate its utility on real robotic tasks, where it learns meaningful rewards with minimal human input.
arXiv:2606.16859v1 Announce Type: new Abstract: Photochemical processes govern phenomena ranging from solar energy conversion and atmospheric chemistry to vision and photosynthesis. Accurate simulation of these processes requires modeling excited-state potential energy surfaces, often involving chemical reactions, tasks that remain computationally prohibitive for extended systems and long timescales using traditional \textit{ab initio} methods. Machine learning interatomic potentials have revolutionized ground-state simulations, but their extension to excited states faces fundamental challenges: standard architectures assume energy extensivity, an assumption that fails for excited states. Here, we present a size-intensive machine-learning framework for excited-state dynamics based on \textit{extremal pooling} of predicted atomic HOMO and LUMO contributions. Trained exclusively on excitations energies and forces, the architecture learns interpretable atomic-level contributions that encode physical information on the extent of electron localization. We demonstrate this framework on the photoexcited solvated electron in liquid water, a paradigmatic problem in radiation chemistry leading to competing pathways involving both hydrogen-atom dissociation and proton-coupled electron transfer. The model not only reproduces the relevant chain of reactions and product species that form during excitation, but also allows one to explicitly study the dynamics of the solvated electron in quantitative agreement with previously reported Restricted Open-Shell Kohn-Sham calculations, while enabling excited-state simulations of periodic systems at length and time scales inaccessible to the reference electronic-structure method. This work establishes a general strategy for machine learning-driven excited-state dynamics applicable to diverse photochemical systems, from molecular chromophores in solution to extended condensed-phase systems.
arXiv:2606.16868v1 Announce Type: new Abstract: While federated learning (FL) enables collaborative medical image segmentation without centralizing sensitive data, real-world deployment is frequently complicated by cross-site label imperfections such as contour disagreement, missing or additional structures, and confused labels. Federated noisy label learning (FNLL) aims to mitigate these effects, yet remains underused in practice as existing evidence is largely based on synthetic noise, simplified settings, and limited real-world noisy evaluation. We address this gap by introducing a benchmark suite that combines diverse real-world noisy datasets, deployment-relevant client-noise scenarios, and label-noise-targeted evaluation to support systematic FNLL assessment and informed method selection. The suite combines curated real-world noisy medical image segmentation datasets from diverse sources with a comprehensive federated segmentation framework including various client-noise scenarios and noise-targeted evaluation. The presented suite provides a realistic and discriminative basis for FNLL evaluation in medical image segmentation and establishes a reusable foundation for fair benchmarking, dataset-specific label-noise characterization, and future method development under realistic federated settings. Code is available at https://github.com/MIC-DKFZ/FedSegNoiseBench.
arXiv:2606.16870v1 Announce Type: new Abstract: Realistic visual simulation of food manipulation requires accurate material parameters, yet these are difficult to measure directly and vary across the heterogeneous regions of a single food item. We address the inverse problem of estimating material parameters from a target description of fracture behavior in a non-differentiable continuum damage mechanics simulator. Using orange peeling as a test case, we train a neural surrogate on 2,000 forward simulations and compare Covariance Matrix Adaptation Evolution Strategy (CMA-ES, a gradient-free evolutionary optimizer) with Proximal Policy Optimization (PPO, a reinforcement learning algorithm) across the original 9-dimensional parameter space and two learned 4-dimensional latent representations. Since different oranges have different material properties, a practical inverse system must handle arbitrary targets without retraining. We train a goal-conditioned PPO policy that learns a general inverse mapping: given any target description of peeling behavior, the policy produces a material parameter estimate in a single forward pass (8 surrogate evaluations, approximately 10ms). Operating in a normalizing flow latent space with a shared surrogate evaluator, the goal-conditioned policy achieves 0.642 actual recovery when validated through the simulator, outperforming the original parameter space by 23%. A warm-start extension that initializes CMA-ES refinement from the policy's output further improves recovery to 0.828 with 540 evaluations. These findings provide a practical framework for inverse food physics and lay groundwork for vision-driven material identification from video observations of food manipulation.
arXiv:2606.16871v1 Announce Type: new Abstract: AI agents must be evaluated as behavioral systems, not as isolated response generators. They reason across turns, call tools, preserve context, follow policies, and act under uncertainty. Existing methods provide useful but fragmented signals: benchmarks measure fixed capabilities, Human-in-the-Loop review preserves expert judgment but does not scale easily, LLM-as-judge methods depend on evaluator design, red teaming is often episodic, and trace auditing requires explicit evidence rules. This paper introduces Human-on-the-Bridge (HOB), a scalable evaluation paradigm for agentic AI. HOB places human expertise upstream, where experts curate reusable evaluation intelligence before testing begins, including domain context, Red-Team Traps, Juror Personas, scoring guidelines, audit rules, and fallback policies. ProofAgent Harness then executes this curated intelligence repeatedly through multi-turn adversarial evaluations, trace capture, multi-juror scoring, and evidence-linked reporting. We evaluate HOB through symmetric and cost-efficient asymmetric settings across frontier LLM-based agents and Harness LLM tiers. The study covers 23,500 agent turns and produces evidence-linked findings across finance, healthcare, and code generation. The results show that HOB can amplify evaluation quality without requiring equally large evaluator models, allowing smaller Harness LLMs to challenge agents built on frontier LLM backbones. The evaluation surfaces failures often missed by static benchmarks and single-evaluator scoring, including phantom tool-call claims, missing mandatory tool calls, policy drift, manipulation paths, and safe but non-resolving refusals. These findings support HOB as a paradigm for scaling human-curated evaluation intelligence, where expert judgment is encoded upfront and reused across repeated agent evaluations rather than applied manually inside every run.
arXiv:2606.16874v1 Announce Type: new Abstract: Online scam behavior is inherently multi-stage, and the lifecycle includes temporally ordered rails and events rather than isolated signals. Existing works analyze characteristics of scam types and rails, but they do not track scam trends across years. Moreover, the work on the relations between rails is hampered due to the lack of open-source datasets with annotations and coverage of different scam types. To address these gaps, we build a dataset to analyze the yearly trend of scam characteristics and rail paths using Reddit self-disclosure narratives from 2023 to 2025. We collect 21,304 posts from scam-related subreddits with at least one rail among identity, communication, platform, and payment for trend analysis by heuristic annotation. Then, we label 1,800 posts containing explicit or recoverable scam chains by an LLM-assisted method for scam path analysis. The method is evaluated with human annotation. Lastly, we run a topic model on the comments of the posts to analyze the community support behavior. The results reveal that scam processes are predominantly multi-rail. Across years, different scam types and rail components dominate. Different scam types vary systematically in path complexity. Reddit support behaviors have become more detailed over time. This work supports synthetic scam chain data simulation and AI-related scam risk assessment, though findings may not generalise to other platforms.
arXiv:2606.16876v1 Announce Type: new Abstract: Adaptive torque prediction in dynamic exoskeleton scenarios requires expensive motion capture systems, which are infeasible in complex outdoor environments. Trajectory prediction has emerged as one of the effective approaches to address such an issue. However, the core challenges of exoskeleton trajectory prediction are twofold: establishing the mapping from multi-modal features to trajectory information; constructing the mapping from trajectory to torque. For the former, most existing methods perform only single-step prediction and neglect inter-subject trajectory variability, thereby limiting the trajectory optimization space and prediction generalization. To address this, this paper proposes a fast flow matching method that enables accurate trajectory prediction and better generalization for real-time performance, where trajectory generation errors and encoded observations are used to guide the training direction. For the second challenge, due to the high dynamics of the human-robot system and the strong coupling between perception and control, simple control methods struggle to achieve efficient assistance based on the predicted trajectory. This paper utilizes model predictive control and designs a novel optimization objective to optimize torque, ensuring the exoskeleton achieves comfortable and robust assistance. By integrating the above two components, the unified policy, denoted as ExoTraj, is developed to enable adaptive assistance in complex outdoor scenarios without high data acquisition cost. Experimental results show that compared to traditional methods, ExoTraj reduces cross-subject prediction error by 14.0% during the online phase and maintains robustness against external noise. Relative to the zero torque condition, ExoTraj decreases metabolic rate by 11.5-24.4%, heart rate by 1.7-19.5%, and peak muscle activation levels by 10.9-41.3%, respectively.
arXiv:2606.16881v1 Announce Type: new Abstract: We introduce a data-efficient distributed Simultaneous Localization and Mapping (SLAM) framework designed for a team of robots equipped with LiDAR, cameras, and inertial sensors. Our framework uses scene graph matching to identify inter-robot measurement constraints. Unlike prior approaches that rely on feature-level matching, our framework is the first to perform scene graph matching using only object labels and centroids. Our approach constructs a scene graph by using fused RGB-LiDAR point clouds to generate both a semantically segmented point cloud layer, and a layer of discrete bounded objects, to accompany estimated robot trajectories. Scene graph matching is performed collaboratively through exchanging and matching object data with neighboring robots. To maximize communication efficiency, we utilize a multi-step data exchange and optimization process. We demonstrate the effectiveness and efficiency of our approach using both simulation and real-world datasets collected by legged robots in indoor and outdoor environments.
arXiv:2606.16888v1 Announce Type: new Abstract: Learning from Demonstration (LfD) enables intuitive robot skill acquisition by allowing robots to learn directly from human task demonstrations. However, current methods often fail to address the fact that due to suboptimal and inconsistent human behavior, the quality of the demonstration can vary within each demonstration. Therefore, we introduce LOPAL (LOcal Performance-aware Active Learning), an active learning approach that leverages this local demonstration quality information. Our approach consists of two synergistic components. First, a local performance-driven LfD method uses a Gaussian Mixture Model (GMM) to encode both the demonstrated trajectories and their associated local quality assessments. This enables the generation of trajectories that outperform the imperfect demonstrations by utilizing complementary local data of high performance. Second, active data acquisition allows to improve beyond the imperfect demonstrations by collecting additional informative samples. In areas missing good data, the user is actively requested to provide corrections through a shared autonomy (SA) mechanism, while the robot autonomously executes the learned behavior. The efficacy of LOPAL was validated in both a simulation and a real-world experiment. The results from a real-world pipe inspection task showed that the proposed approach can achieve up to 27.31 % improvement in task performance while also reducing the effort required to collect the demonstrations.
arXiv:2606.16889v1 Announce Type: new Abstract: As the carbon cost of manufacturing and operating semiconductor devices has come into sharper focus, sustainability has gradually emerged as a new system architecture design metric. Like power and performance modeling tools, enabling sustainability-aware silicon systems design and optimizations will require a new generation of electronic design automation and architectural modeling tools. Towards this end, we present an update to the Architecture Carbon Tool v3 (ACT3) which aims to provide an extensible and customizable modeling platform for research and advanced development to pave the path towards sustainability-aware architectural design space exploration. Compared to previous versions of ACT, ACT3 provides significantly richer modeling capabilities, enhanced collateral and analysis telemetry, and first order design space exploration capabilities. This technical brief provides an overview of these expanded capabilities and illustrates ACT3's basic utility across several case studies. Finally, we identify opportunities for research and development where we expect the research community can contribute towards continuing to improve sustainability modeling and design methodology for silicon systems.
arXiv:2606.16890v1 Announce Type: new Abstract: Aggregate accuracy benchmarks conceal a systematic structure in how large language models fail at electronic health record (EHR) question answering: questions requiring more inferential steps produce disproportionately more errors. Motivated by theoretical results on transformer compositionality limits, we introduce a pre-specified hop-count taxonomy -- the number of distinct reasoning steps required to answer a clinical question from an EHR -- as a principled predictor of model failure. We annotate 313 clinician-generated MedAlign EHR question-answer pairs across four hop levels and evaluate 301 questions in a within-model ablation (claude-sonnet-4-6, zero-shot vs. extended thinking) and cross-architecture replications (gpt-4o and gpt-5.4-2026-03-05, zero-shot). All three models, spanning two providers and two OpenAI generations (GPT-4 and GPT-5), show monotone accuracy decline with hop count: Claude Sonnet zero-shot falls from 30.6% (hop=1) to 17.6% (hop=4) (Cochran-Armitage z=-2.30, p=0.011; OR per hop 0.72, 95% CI [0.56,0.92], p=0.008); GPT-4o replicates this (37.8% to 14.7%; OR 0.58 [0.45,0.75], p<0.001); and gpt-5.4-2026-03-05 confirms it (37.8% to 23.5%; OR 0.80 [0.66,0.98], p=0.027). A pre-specified context-sufficiency audit shows higher-hop questions are not differentially disadvantaged by EHR truncation (answerability 93-95% at hops 2-4 vs. 79% at hop=1), so the decline reflects compositional reasoning difficulty. Extended thinking did not significantly flatten the accuracy-depth curve across three reasoning conditions, and thinking-token usage scaled with hop count (r=0.31, p<0.0001), consistent with the predicted O(k) computational requirement. Hop count is thus a theory-motivated, cross-architecture predictor of large-language-model error on EHR question answering, with direct implications for deployment risk stratification of clinical AI.
arXiv:2606.16891v1 Announce Type: new Abstract: Federated Learning is rapidly evolving beyond the exchange of traditional model weights and gradients, yet existing definitions fail to capture the full scope of modern payloads like synthetic data and federated analytics. This paper addresses the gap by proposing a formal mathematical definition of a federated message that accounts for both utility and privacy. We introduce a taxonomy that organizes these exchanges into three categories: model structures, statistical summaries, and data-conditioned representations. By evaluating these groups based on computational demands, communication costs, and privacy risks, we provide a clearer understanding of the trade-offs involved in decentralized training. Our review of 202 recent publications highlights a significant shift since 2021 toward diverse messaging paradigms, signaling a move away from standard deep learning updates toward more specialized information sharing. This framework provides a structured path for future research to optimize federated systems for varying hardware and security requirements.
arXiv:2606.16892v1 Announce Type: new Abstract: Gaussian networks are degree-four symmetric interconnection networks defined over residue classes of Gaussian integers. Earlier work showed that when the generator $\alpha=a+bi$ satisfies $\gcd(a,b)=1$, the real and imaginary dimensions directly form two edge-disjoint Hamiltonian cycles. A later construction extended the result to the non-coprime case $\gcd(a,b)=d>1$, but its proof used long node-sequence tables and separate odd/even cases for $d$. This paper gives a unified closed-form construction that covers both $d=1$ and $d>1$, and also covers both odd and even $d$, without separate case tables. In the rectangular representation with $d$ rows and $r=(a^2+b^2)/d$ columns, the construction uses a constant-time local switch rule for each $q=1,2,\ldots,d-1$ at column $a_q=q-1$. Each switch removes two horizontal edges and inserts two vertical edges. The switched horizontal structure forms the first Hamiltonian cycle, while its edge-complement in the Gaussian network forms the second Hamiltonian cycle. Thus, the full edge set is partitioned into two edge-disjoint Hamiltonian cycles. The construction requires $O(d)$ switch-generation time and $O(N)$ time to list the two cycles, where $N=a^2+b^2$. Exhaustive validation for all $1\leq a\leq b\leq 100$, excluding only the degenerate $N=2$ network, and large-scale validation up to $N=3{,}250{,}000$ confirm the construction.