arXiv:2607.02576v1 Announce Type: cross Abstract: Parameter-Efficient Fine-Tuning (PEFT) commonly adapts pretrained weights through low-rank updates, and recent methods further exploit the singular value decomposition (SVD) of the base weight for initialization or subspace selection. However, these methods do not explicitly preserve the coupled geometry between the pretrained left and right singular bases. Motivated by recent minimum-perturbation theory, which shows that stable finetuning follows a coherent SVD rotation in which a single orthogonal $Q$ acts on both the left singular basis $U_0$ and the right singular basis $V_0$, we prove a per-slice analogue: each row slice of $W_0$ can be adapted by a shared orthogonal rotation $Q_i$ on its left basis $U_i$ and right basis $V_i$ together with a diagonal spectrum shift. We implement this form as CORA (Coherent Orthogonal Rotation Adaptation), which applies per-slice orthogonal rotations and a per-layer diagonal scale to the rank-$r$ SVD truncation of $W_0$. CORA uses $\tfrac{1}{2}m(r{-}1)$ trainable parameters per linear layer, about $4{\times}$ fewer than LoRA at the same rank. CORA outperforms LoRA, DoRA, PiSSA, and MiLoRA on commonsense reasoning and code generation while using about $8{\times}$ fewer parameters.
Science Journals
arXiv:2607.02622v1 Announce Type: cross Abstract: Multiplex CRISPR-Cas9 gene editing requires selecting one guide RNA per target gene subject to cross-gene interactions: a constrained combinatorial problem that can be formulated as a Quadratic Unconstrained Binary Optimization (QUBO) and solved via the Quantum Approximate Optimization Algorithm (QAOA). The one-hot per-gene constraint is conventionally enforced by adding quadratic penalty terms to the cost Hamiltonian, but penalty coefficient selection is heuristic and penalties amplify hardware noise. An alternative is to enforce the constraint structurally via the XY-mixer, which preserves feasibility by construction. We present COMET, a systematic comparison of penalty-based and XY-mixer QAOA on a three-gene, twelve-qubit multiplex editing instance targeting the immune-checkpoint genes PDCD1, LAG3, and HAVCR2. In simulation, the XY-mixer exceeds 95% probability of the optimum by QAOA depth p=3, while three penalty variants spanning an order of magnitude in penalty coefficient remain below 6% at every depth. On IBM's ibm_kingston (Heron r2) processor, the XY-mixer's simulator-hardware energy gap stays within |0.8| across all depths, while the worst-tuned penalty variant's gap reaches +53.9. We provide an honest account of where the structural guarantee partially breaks under gate-level noise. The twelve-qubit instance is classically trivial; our contribution is a methodological comparison of constraint-enforcement strategies in a biologically motivated domain, with real-hardware validation.
arXiv:2607.05101v1 Announce Type: new Abstract: The application of machine learning-based predictive algorithms to Anti-Money Laundering (AML) has grown rapidly, driven by the vast volume of financial transaction data available to banks. These algorithms are typically trained not only on transactional data but also on sensitive client information, which may raise fairness concerns. Despite this, AML detection systems remain largely underexplored from a fairness perspective, even though deeper analytical methods based on counterfactuals are now available. Such techniques enable the decomposition of the direct and indirect effects of potentially sensitive features on model predictions, thereby supporting the evaluation of whether their influence is acceptable from a fairness perspective. Closing this gap, we consider the synthetic IBM AMLSim transaction dataset and construct additional features of the country of an account and its average behaviour. This improves the predictive performance of diverse machine learning models, ranging from baseline decision trees to state-of-the-art graph neural networks. We assess the potential unfairness associated with these features through a counterfactual, path-specific effect analysis. This reveals that fairness violations tend to be more pronounced for models whose predictive performance benefits the most from the extended features. Such a finding highlights a concrete instance of the trade-off between predictive accuracy and fairness in AML applications, thus underscoring the urgency of a systematic fairness analysis in such critical domains.
arXiv:2506.23281v3 Announce Type: replace Abstract: Random testing has proven to be an effective technique for compiler validation. However, the debugging of bugs identified through random testing presents a significant challenge due to the frequent occurrence of duplicate test programs that expose identical compiler bugs. The process to identify duplicates is a practical research problem known as bug deduplication. Prior methodologies for compiler bug deduplication primarily rely on program analysis to extract bug-related features for duplicate identification, which can result in substantial computational overhead and limited generalizability. This paper investigates the feasibility of employing bisection, a standard debugging procedure largely overlooked in prior research on compiler bug deduplication, for this purpose. Our study demonstrates that the utilization of bisection to locate failure-inducing commits provides a valuable criterion for deduplication, albeit one that requires supplementary techniques for more accurate identification. Building on these results, we introduce BugLens, a novel deduplication method that primarily uses bisection, enhanced by the identification of bug-triggering optimizations to minimize false negatives. Empirical evaluations conducted on five real-world datasets demonstrate that BugLens significantly outperforms the state-of-the-art analysis-based methodologies Tamer and D3 by saving an average of 33.56% and 10.68% human effort to identify the same number of distinct bugs. Given the inherent simplicity and generalizability of bisection, it presents a highly practical solution for compiler bug deduplication in real-world applications.
arXiv:2607.02537v1 Announce Type: cross Abstract: Jamming and spoofing pose significant threats to wireless and satellite navigation by disrupting radio-frequency (RF) signals and compromising availability and integrity. Robust RF interference direction finding through angle-of-arrival (AoA) estimation is therefore essential for detecting and localizing anomalous signals. Although data-driven methods perform well under line-of-sight (LoS) conditions, their performance degrades in practical environments due to non-line-of-sight (NLoS) multipath propagation. In this work, we propose a hybrid learning framework that incorporates physics-informed constraints into deep neural networks to improve the robustness of AoA estimation. A neural network is trained to estimate the azimuth and elevation of incoming signals received by a four-element antenna array, while a physics-informed loss enforces consistency between the predicted angles and inter-antenna phase differences under a plane-wave model. We further introduce a latent-space classifier to distinguish LoS from NLoS samples. Since inter-antenna phase differences under LoS propagation exhibit domain-invariant structure across environments, the physics-based loss is applied only to LoS samples, promoting physically consistent and domain-invariant representations without over-constraining the model in NLoS scenarios. In addition, domain-incremental learning (DIL) across NLoS environments with varying scatterer distributions improves cross-domain generalization. Evaluations on real-world datasets show that the proposed method reduces AoA estimation error by up to 6{\deg} in low-exemplar settings compared with DIL baselines.
arXiv:2607.05120v1 Announce Type: new Abstract: AI agents act on behalf of user prompts, consuming external data and taking actions based on the agent context. Prior research on AI agent security has primarily focused on indirect prompt injection (IPI). Its most well-studied category is instruction injection, where attacker-controlled untrusted data is interpreted as an instruction. In response, many mitigations have been proposed to prevent instruction injection attacks. In this paper, we introduce a new category of IPI, agent data injection attacks (ADI). ADI injects malicious data disguised as trusted data, such as security-critical metadata (e.g., resource identifiers or data origins) or agent context data (e.g., tool call and response formats). As a result, agents unknowingly execute unintended actions based on attacker-controlled data. ADI has similar attack impacts as instruction injection attacks, because it causes agents to misbehave and execute unintended actions. Despite the similar impact, ADI remains underexplored and easily bypasses existing IPI defenses. We found several critical vulnerabilities in real-world agents that allow an attacker to launch various attacks: arbitrary click attacks on web agents (Claude in Chrome, Antigravity, and Nanobrowser), and remote code execution and supply-chain attacks on coding agents (Claude Code, Codex, and Gemini CLI). We evaluate ADI vulnerabilities across off-the-shelf models and AI agents, and find that ADI is effective in both standalone LLMs and AI agent settings. ADI exposes a critical gap in agent security, signifying that current AI agents do not employ a fundamental security principle: current agents do not isolate trusted data from untrusted data.
From Failing to Passing: Evolving Natural Language Prompt Optimization Rules for LLM Code Generation
arXiv:2607.05121v1 Announce Type: new Abstract: Large language models are known to be sensitive to prompt formulation. Even minor variations in wording can substantially degrade performance. This sensitivity reveals an opportunity: if prompt phrasing can harm performance, can it be used to improve it? To investigate this question, we introduce a search-based approach that identifies and evolves a set of natural language transformation rules with strong downstream effects on coding performance. We then propose DUALFIX, a staged repair pipeline that combines the evolved transformation rules with execution-feedback repair, addressing both specification-level and implementation-level failures. A key strength of our approach lies in its generality: the evolved rules are error-agnostic, reusable across problems, and transferable across models. We evaluate DUALFIX against execution-feedback repair baselines across three models on two challenging benchmarks, LiveCodeBench and APPS. Our results show that the evolved transformations fix from 10-30% of failing cases, including 12-17% of failures that execution-based repair alone cannot resolve. Overall, DualFix recovers up to 30% of baseline failures and fixes 3-5 times more failing cases than Self-Fix across all evaluated settings. Furthermore, we also show that rules evolved on one model transfer zero-shot to other models, outperforming execution-feedback repair without any re-optimization.
arXiv:2607.05122v1 Announce Type: new Abstract: Vision-language-action (VLA) models enable robot navigation from natural language and visual goals, but remain susceptible to perceptual distractions and ambiguous scene interpretations. This paper presents the first empirical evaluation of visual grounding for VLA navigation policies. We propose a real-time segmentation-based grounding method that highlights traversable areas in green and non-traversable areas in red using SegFormer. Two variants are evaluated: observation-only segmentation and joint observation-goal augmentation. Using OmniVLA on the Grand Tour dataset, we show that visual grounding reduces the mean waypoint error by 27-44% at the farthest waypoint, depending on the instruction length. The benefits are greater for long instructions than for short instructions, and grounding provides little improvement for image goals. Normalized error analysis indicates that grounding primarily acts as a trajectory length regularizer, reducing the predicted path length by 30% without improving per-unit-distance reasoning. Our results indicate that visual grounding offers a simple, computationally inexpensive method to improve VLA navigation without model retraining, although it cannot compensate for missing training signals in out-of-distribution instructions.
arXiv:2607.05126v1 Announce Type: new Abstract: We investigate quantifier-free interpolation properties for several fragments generalising the extensional theory of arrays. Our results include the (general) quantifier-free interpolation properties of combinatory array logic with iterated diffs and the uniform interpolation property of the simple flat array fragment. To our knowledge, these are the first positive quantifier-free interpolation results obtained for theories of arrays featuring expressive specifications over unbounded domains.
arXiv:2512.04073v2 Announce Type: replace Abstract: Stop-and-go waves in traffic flow are captivating collective phenomena with important safety and environmental consequences. While classical theories attribute these oscillations to linear instabilities caused by reaction delays and inertia, this study explores an alternative stochastic perspective. Using a linearly stable car-following model, we show that white Gaussian noise in the measurement of the inter-vehicle distance can destabilise the flow, inducing a phase transition to periodic stop-and-go dynamics via a nonlinear instability mechanism. Furthermore, we demonstrate that a simple linear transformation of the model, which amplifies the system response while introducing a positive acceleration bias, can counteract noise-induced effects and restores the stability of uniform traffic flow. These findings, supported by numerical simulations, aim to provide new insights into the modelling and control of oscillatory traffic dynamics.
arXiv:2607.05129v1 Announce Type: new Abstract: Two-photon fluorescence imaging achieves deep-tissue penetration through long excitation wavelengths and nonlinear excitation confinement. The 1700 nm transparency window is particularly attractive, as it optimally balances tissue scattering and absorption. However, efficient fluorophores for two-photon excitation in this window remain limited. Moreover the weak near-infrared emission of individual emitters, and the low photon detection efficiency, has so far precluded single-particle imaging. Here, we characterize the two-photon excitation properties of chirality-sorted pristine and quantum color center-functionalized single-walled carbon nanotubes under 1700 nm excitation. By measuring and comparing their two-photon action cross-sections, we identify quantum color center-functionalized (6,5) nanotubes emitting at 1140 nm, as the most promising emitter, with an exceptionally large cross-section of (57 \pm 2).103 GM. Leveraging these favorable photophysical properties, we image individual nanotubes under 1700 nm excitation, to our knowledge the first demonstration of single-particle imaging at this wavelength. These results establish quantum color center-functionalized (6,5) nanotube as a strong candidate for long-wavelength two-photon imaging and lay the groundwork for deep-tissue single-particle imaging.
arXiv:2607.05131v1 Announce Type: new Abstract: Among the five primary human senses, tactile is arguably the most fundamental to survival, as it enables the perception of physical contact and interaction in real-world environments. In this paper, we explore two key challenges of integrating tactile sensing into intelligent systems for multimodal reasoning: (i) insufficient modeling of dynamic tactile signals, which restricts reasoning over temporally evolving properties, and (ii) hallucination in tactile foundation models caused by the absence of explicit reasoning mechanisms, leading to unstable real-world inference. To address these challenges, we propose TacReasoner, a dynamic tactile-language framework for interactive reasoning in real-world scenarios. First, TacReasoner incorporates a Dynamic-aware Tactile Encoder to enhance the perception and representation of dynamic tactile signals. More importantly, we introduce TouchCoT-10k, the first tactile chain-of-thought dataset for structured reasoning over tactile inputs. Upon it, we establish DynTac-Bench to systematically evaluate dynamic tactile perception and real-world commonsense reasoning. Experimental results demonstrate that TacReasoner achieves competitive performance against state-of-the-art models across multiple datasets. Notably, despite using only 7B parameters, TacReasoner outperforms the 14B VTV-LLM model on most subtasks, highlighting its effectiveness and efficiency in tactile commonsense reasoning.
arXiv:2507.19801v2 Announce Type: replace-cross Abstract: Young's double slit experiment has often been used to illustrate the concept of complementarity in quantum mechanics. If information can in principle be obtained about the path of the photon, then the visibility of the interference fringes is reduced or even destroyed. This Gedanken experiment discussed by Bohr and Einstein can be realized when the slit is replaced by individual atoms sensitive to the transferred recoil momentum of a photon which "passes through the slit". Early pioneering experiments were done with trapped ions and atom pairs created via photo-dissociation. Recently, it became possible to perform interference experiments with single neutral atoms cooled to the absolute ground state of a harmonic oscillator potential. The slits are now single atoms representing a two-level system, and the excitation in the harmonic oscillator potential is the which-way marker. In this note, we analyze and generalize two recent experiments performed with single atoms and emphasize the different ways they record which-way information.
arXiv:2603.01231v2 Announce Type: replace Abstract: Velocity-model building is a fundamental component of seismic imaging, yet it remains a challenging inverse problem due to limited data coverage, nonlinearity, and the need to integrate heterogeneous information such as well logs. We introduce a unified framework for velocity-model editing and full velocity-model building that combines learned diffusion priors with structurally preconditioned inverse formulations. A diffusion model trained on high-resolution synthetic velocity examples provides a data-driven prior that is exploited through Denoising Diffusion Implicit Model (DDIM) inversion and guided sampling. For localized editing, the diffusion prior is coupled with a structurally preconditioned Tikhonov well-matching inversion, enabling controlled modification of selected regions while preserving global consistency. For full velocity-model building, we formulate a well-matching inverse problem augmented with imaging-based regularization and solve it using conventional least-squares, the proposed DDIM-guided method, and Diffusion Posterior Sampling (DPS). Synthetic experiments demonstrate that diffusion-based approaches recover sharper and more realistic velocity structures than classical inversion. Field-data applications on the Viking Graben dataset confirm robustness under realistic acquisition conditions. An ablation study highlights the critical role of structural slope guidance in inversion performance. Overall, the proposed framework bridges inverse problems and generative modeling, offering a flexible approach for practical seismic imaging workflows.
arXiv:2512.13099v2 Announce Type: replace Abstract: Driven by the ongoing energy transition, shared mobility providers are emerging actors in power systems aiming to shift combustion-based vehicles towards electric ones. Meanwhile Energy Communities are deployed to promote investment in distributed renewable production and enhance the local usage of it. The complementarity in their electrical demand, enhanced by a coordinated operational planning, can help both actors reduce the electricity supply cost. Considering this original collaboration, this paper presents a Mixed-Integer Quadratic Programming problem which jointly optimizes the EC members and EVs flexibility usage to take advantage of the local production. Besides economic benefits comparison, authors analyses the impact of grid tariffs and bi-directional charging on the distribution network. Results from a Belgian mobility case study show that coordination can help reducing the yearly cost up to 15.6% compared to their stand-alone situation and that it may reduce by 30.8% the stress on the substation transformer when subject to peak penalties from the grid operator.
arXiv:2607.05177v1 Announce Type: new Abstract: Workforce scheduling is an NP-hard combinatorial optimization problem requiring simultaneous satisfaction of labor regulations, coverage requirements, employee preferences and operational objectives. Existing CP formulations typically model simplified instances with 6-12 constraints at shift-level granularity and critically lack explicit support for: mandatory break scheduling with midpoint placement control; acuity weighted workload equity; sub-shift temporal granularity enabling demand-driven staffing; inter-week schedule stability; and cross-midnight shift patterns common in 24-hour operations. This paper presents CP-WSP: a declarative CP-SAT framework enforcing 14 hard constraints as mathematically inviolable requirements (zero regulatory violations by construction) while optimizing 15 soft objectives through a unified weighted penalty function -- all configurable via a JSON specification with no code changes required. Key contributions include: a shift-window variable decomposition enabling mandatory break scheduling with centrality control; acuity-weighted workload equity; multi-granularity temporal resolution from 30 minutes to 2 hours; inter-week schedule stability; a grid-offset preprocessing technique for cross-midnight shifts; and a reproducible 36-configuration benchmark suite for community comparison. Evaluated on INRC-II benchmarks at both hourly and shift-level granularity and on 36 synthetic configurations.
arXiv:2607.05197v1 Announce Type: new Abstract: Iterative repair loops have become a core design pattern in LLM-based software engineering systems. These workflows repeatedly generate, validate, and repair artifacts using feedback such as compiler errors or test failures. Despite their widespread use, the impact of repair-loop iteration limits remains poorly understood, as most prior work adopts fixed, often arbitrary, repair budgets. We study repair-loop effectiveness across multiple software engineering tasks, including code generation, test generation, and code translation. Across several representative workflows, datasets, and contemporary low-cost LLMs, we observe a consistent pattern of diminishing returns: the first three to four repair iterations account for most achievable gains, while later iterations contribute only marginal improvements. We further find that repair behavior is influenced more strongly by workflow orchestration and feedback design than by the underlying model itself. These results suggest that repair budgets should be treated as an explicit experimental variable, as they directly affect evaluation outcomes, computational cost, runtime, and reproducibility in LLM-based software engineering research.
arXiv:2607.05198v1 Announce Type: new Abstract: Minimum Bayes Risk (MBR) decoding yields more robust and higher-quality text generation than maximum a posteriori (MAP) decoding by selecting hypotheses that maximize expected utility over sampled pseudo-references. However, there exists a discrepancy in the design: hypothesis selection calculates expected utility scores conditioned on given pseudo-references, while commonly used evaluation metrics, e.g., BLEU and COMET, are asymmetric. Therefore, it is important to consider both hypothesis-to-reference and reference-to-hypothesis directional effects. In this study, we introduce a noisy channel decomposition of MBR decoding that naturally incorporates bidirectional effects to account for these asymmetries. We decompose MBR decoding into four interacting components: hypothesis-to-reference likelihood, reference-to-hypothesis likelihood, hypothesis prior, and reference prior. This decomposition provides a unified interpretation of existing MBR variants and enables metric- and task-specific interpretability by isolating the contribution of each channel. Our comprehensive analysis reveals that channel-wise contributions exhibit distinct characteristics across metrics while remaining consistent across tasks, and suggests that appropriate channel weighting may lead to improvements over original MBR decoding.
arXiv:2607.05272v1 Announce Type: new Abstract: Inference serving systems must balance throughput and latency under bursty, heterogeneous workloads, yet the industry standard remains static batching policies that require manual tuning and cannot adapt to shifting traffic. We investigate whether reinforcement learning (RL) can learn adaptive batching and routing policies that outperform these heuristics, training REINFORCE and PPO agents on a discrete-event simulator validated against queuing theory and production traces (Azure Functions, BurstGPT). We formulate the problem as an MDP over queue state, request type and GPU availability, evaluating across standard Poisson traffic, extreme bursts, real-world traces and heterogeneous multi-GPU routing. Our central finding is a clear boundary condition for RL's value in systems problems. In single-GPU settings, a well-tuned static batching policy is already near-optimal under Poisson-like arrivals and RL offers only marginal gains (+0.1% to +1.0%). In multi-GPU heterogeneous routing, however, where fast and slow requests compete for shared resources, the agent discovers a workload-segregation policy that eliminates Head-of-Line blocking, yielding a 3.5x (348%) improvement over Round-Robin and a 48% improvement over the strongest heuristic baseline (Shortest-Queue), with 60% higher throughput and 25% lower latency while respecting SLA constraints. The policy generalizes to unseen bursty and real-world traffic despite training only on synthetic Poisson arrivals and an attention-augmented policy network converges roughly 20% faster than an MLP baseline. These results suggest RL's advantage over engineered heuristics concentrates in combinatorial, multi-resource decisions rather than single-resource temporal scheduling, a practical distinction for deciding where learned policies justify their engineering cost in production inference infrastructure.
arXiv:2604.25581v2 Announce Type: replace-cross Abstract: We investigate self-consistent vertex corrections to the $GW$ self-energy for ionization potentials (IPs) and electron affinities (EAs). We benchmark IPs against $\Delta$CCSD(T) references in the $G_0W_0\Gamma$29 and GW100 sets and compare GW100 EAs with EOM-CCSD references. Because many anions are metastable, these addition energies are model quantities and should not be interpreted as experimental EAs. Tensor hypercontraction (THC) of the Coulomb integrals enables efficient self-consistent $GW\Gamma_{\Sigma}$ implementations, where vertex corrections are included only in the self-energy. We establish a hierarchy of vertex-corrected self-energies relative to sc$GW$, ordered from least to most negative as SOX $>$ SOSEX $>$ G3W2 $>$ 2SOSEX $>$ sc$GW$. Equivalently, the absolute magnitude increases along this sequence. This trend follows an effective-screening picture, in which increasing screening progressively reduces exchange contributions. Static and dynamic variants show consistent differences due to the frequency dependence of the screened interaction. Across all methods, vertex corrections act as an approximately frequency-uniform self-energy renormalization rather than altering its spectral structure. In terms of accuracy, sc$GW\Gamma_{\Sigma}$ does not uniformly improve IPs or the EA model quantities over sc$GW$. For IPs, SOX and SOSEX usually degrade performance, whereas 2SOSEX and G3W2 remain close to sc$GW$, with only marginal improvements for selected systems at higher cost. Although the tested variants reduce EA MAEs, this should not be interpreted as a general improvement for physical anions because many nominal EA states are metastable. These results indicate that generic vertex insertions are insufficient to outperform sc$GW$; systematic improvements require designed diagrammatic approximations combined with efficient tensor factorization.
arXiv:2604.02923v4 Announce Type: replace Abstract: Large Language Models (LLMs) have demonstrated advanced capabilities but often suffer from factual inaccuracies (hallucinations) and systematic biases. These issues, sometimes amplified in specific architectures like Mixture-of-Experts (MoE) which motivate our work, pose risks for reliable deployment. To address these challenges, we propose the Council Mode, a multi-agent consensus framework. Our approach dispatches queries to multiple heterogeneous frontier LLMs in parallel and synthesizes their outputs using a dedicated consensus model. The pipeline consists of three phases: an intelligent triage for query complexity, parallel generation across diverse models, and a structured synthesis that identifies agreement, disagreement, and unique findings. In our evaluation, conducted under controlled no-web settings, the Council Mode achieved a 41.7% relative reduction in hallucination rates on a 1,200-sample HaluEval subset and a 7.5-point improvement on TruthfulQA compared to the top-performing individual model. On our curated MDR-500 multi-domain reasoning benchmark, the Council Mode achieved a Quality Score of 95.4%, representing a 9.2-point improvement over the best individual model. The framework also exhibited lower measured bias variance under our rubric-based evaluation protocol. We provide a cost-effectiveness analysis showing that the framework incurs a 4.2x token-cost overhead, making it most suitable for accuracy-prioritized applications where the cost of errors exceeds the added inference cost. These findings suggest that structured multi-agent consensus is a promising direction for enhancing the reliability and factual grounding of LLM-generated content.
arXiv:2607.05073v1 Announce Type: cross Abstract: This paper studies a non-separable composite $\ell_0$-$\ell_2$ regularization model that simultaneously enforces sparsity and smoothness for inverse problems. The $\ell_0$ norm induces inherent nonconvexity and nonsmoothness, while linear transformations further introduce nonseparability, making the problem computationally challenging to solve. The existing inexact augmented Lagrangian method suffers from high computational complexity and unstable convergence. To overcome these difficulties, we develop two novel augmented Lagrangian algorithms with exact multipliers, designed respectively for the full row-rank case and the general matrix case, where all subproblems are globally optimized via closed-form solutions. Furthermore, we prove linear convergence of the proposed method when the transformation matrix is full row rank. In the general setting, all accumulation points of the generated sequence are KKT points for the original problem. Numerical experiments on synthetic data, trend filtering, and image smoothing demonstrate the superior efficiency and accuracy of the proposed methods over the existing method, confirming our theoretical analysis.
arXiv:2602.24044v2 Announce Type: replace Abstract: Large Language Model (LLM) adapters enable low-cost model specialization, but introduce complex caching and scheduling challenges in distributed serving systems where hundreds of adapters must be hosted concurrently. While prior work has largely focused on latency and throughput optimization, minimizing GPU resource requirements through near-peak utilization remains largely underexplored. This paper presents a data-driven pipeline that, for a given workload, computes an adapter placement that serves the workload with the minimum number of GPUs while avoiding request starvation and GPU memory errors. To that end, the approach identifies the maximum feasible throughput attainable on each GPU by leveraging accurate performance predictions learned from real serving behavior. The proposed pipeline integrates three components: (i) a Digital Twin (DT) tailored to LLM-adapter serving, (ii) a distilled machine learning (ML) model trained on DT-generated data, and (iii) a greedy placement algorithm that exploits ML-based performance estimates to maximize GPU efficiency. The DT emulates real system dynamics with high fidelity, achieving below 5% throughput estimation error while executing up to 90x faster than full LLM benchmarking across both predictable and unpredictable workloads. The learned ML models further accelerate performance estimation with marginal accuracy degradation, enabling scalable optimization. Experimental results demonstrate that the pipeline substantially improves GPU efficiency, reducing the number of GPUs required to sustain target workloads by 60\% on average across the evaluated scenarios. Beyond GPU efficiency, the pipeline can be adapted to alternative objectives, such as latency minimization, highlighting its versatility for future large-scale LLM serving infrastructures.
arXiv:2607.03698v1 Announce Type: cross Abstract: Gold nanonecklace networks are promising platforms for single-electron switching, chemical sensing, and biogating devices because of their nonlinear current--voltage ($I$--$V$) characteristics arising from collective Coulomb-blockade transport. However, the mechanisms governing this macroscopic behavior remain poorly understood because experimental measurements are generally limited to the network topology and global $I$--$V$ response. To address this, we developed a graph-based Kirchhoff framework that represents a self-assembled nanonecklace network as a graph, with nodes corresponding to junctions between necklace segments and edges to the conducting segments themselves. The solver returns the active nodes, conducting subgraph, nodal potentials, and edge currents at each applied bias, while allowing the activation-voltage statistics, network density, and structural topology to be varied independently. The model reproduces the experimentally observed non-Ohmic response, $I \propto (V-V_T)^{\zeta}$, and shows that this behavior emerges from the collective, staggered activation of threshold junctions and voltage-driven percolation of the conducting subgraph. Independent parameter sweeps reveal that the mean activation voltage shifts the threshold $V_T$ while leaving $\zeta$ nearly unchanged, increasing network density raises $\zeta$ from approximately 1.9 to 3.1 and enhances current, and topology controls the response even at fixed density and node characteristics. These trends agree qualitatively with experimental observations and establish the model as a design tool for engineering collective transport in self-assembled nanonecklace devices.
arXiv:2603.01878v2 Announce Type: replace Abstract: Recent advances in generative AI have made synthetic Computed Tomography (CT) images increasingly realistic, enabling promising applications in medical data augmentation while raising serious concerns about clinical safety and data trustworthiness. Detecting AI-generated CT images remains challenging for two key reasons: existing benchmarks cover only limited generation sources, and many detectors are adapted from natural-image forensics without explicitly modeling CT-specific imaging properties. In this paper, we introduce CTForensics, a dataset for detecting AI-generated CT images. CTForensics contains 75,990 2D CT images, including a dedicated test benchmark of 29,990 balanced authentic and generated samples from ten representative CT generative models spanning GAN-based and diffusion-based paradigms. We further propose the Enhanced Spatial-Frequency CT Forgery Detector (ESF-CTFD), a CT-oriented CNN framework built around a Wavelet-Enhanced Central Stem, Multi-Scale Spatial Aggregation, and a Frequency-Aware Prediction Block. The Wavelet-Enhanced Central Stem enhances local intensity correlations and high-frequency residuals, Multi-Scale Spatial Aggregation aligns anatomical features across resolutions with lightweight residual units, and the Frequency-Aware Prediction Block models global spectral artifacts. Extensive experiments on CTForensics show that ESF-CTFD achieves 96.01% mAcc and 99.96% mAP, outperforming existing methods and maintaining strong robustness under realistic perturbations with only a 0.99% average drop. Codes will be available at https://github.com/liyih/CTForensics.