arXiv:2407.10883v2 Announce Type: replace Abstract: Data sharing is increasingly essential for digital government and data-driven innovation, yet many public organizations remain reluctant to make their data openly available. While prior research has examined factors influencing open data adoption, little theoretical work explores why resistance persists within public agencies. This study develops an Innovation Resistance Theory (IRT) model tailored to government data sharing to identify predictors of organizational resistance. An initial model was derived from literature and refined through interviews with 21 public organizations across six European countries. The resulting IRT4DS model identifies 39 barriers spanning usage, value, risk, tradition, and image dimensions, and 23 countermeasures mapped to the most critical barriers and the actors responsible for addressing them. By extending IRT into the context of governmental data sharing, the study advances theoretical understanding of why public data often remains closed and provides actionable guidance for policymakers seeking to design enabling data ecosystems and reduce structural and cultural barriers to OGD adoption.
Science Journals
arXiv:2407.13303v2 Announce Type: replace Abstract: Conventional large-scale indoor localization based on Wi-Fi RSSI fingerprinting faces issues of time-consuming and labor-intensive labeled data collection, limited generalization of a model trained under a supervised learning (SL) framework due to its inability to leverage unlabeled data, and model performance degradation in dynamic scenarios with environmental variations. To address those challenging issues, we propose a comprehensive semi-supervised learning (SSL) framework for a deep neural network (DNN) localization model based on the Mean Teacher, which incorporates access point selection, model pre-training/cloning, and batch-level noise injection. The proposed SSL framework can not only efficiently use hybrid labeled/unlabeled databases for static training of a model during the offline phase, but also exploit unlabeled fingerprints from users of the indoor localization system deployed in the field for continuous retraining of the model during the online phase. We base the proposed SSL framework on the Mean Teacher because it can generate more stable target labels through an exponential moving average of model weights without incurring the high computational complexity of the Pi-Model and with better scalability for online learning than Temporal Ensembling, making it an optimal choice that strikes the right balance between performance and computational complexity in large-scale indoor localization. With the UJIIndoorLoc database, the proposed SSL framework reduces the mean 3D errors of the CNNLoc and SIMO-DNN models by 7.403% and 7.748%, respectively, compared with those under the conventional SL framework; with the XJTLU dynamic database, the maximum reduction in mean 2D error reaches up to 49.227% under a dynamic training scenario, demonstrating the substantial performance improvement achieved by the proposed SSL framework.
arXiv:2408.04207v3 Announce Type: replace Abstract: The field of multiobjective evolutionary algorithms (MOEAs) often emphasizes its popularity for optimization problems with conflicting objectives. However, it is still theoretically unknown how MOEAs perform compared with typical approaches outside this field. This paper conducts such a systematic theoretical comparison on problem classes with different degrees of conflict. With OneMaxMin$_k$ depicting $k\in[0..n]$ degrees of conflict, we show the difficulties of two typical non-MOEA approaches, the scalarization (weighted-sum) and {the} $\epsilon-$constraint approach. We prove that for any set of weights, the set of optima formed by {the} scalarization approach cannot cover its full Pareto front for $k>2$. Although constrained problems constructed from $\epsilon-$constraint approach ensure the full coverage, general ways (via exterior or nonparameter penalty functions) to solve these constrained problems encounter difficulties. The nonparameter penalty function way cannot guarantee the full coverage, and the exterior way covers the Pareto front with expected $O(\max\{k,1\}n\ln n)$ number of function evaluations, but only with careful settings of $\epsilon$ and $r$ ($r>1/(\epsilon+1-\lceil \epsilon \rceil)$). In contrast, MOEAs efficiently solve OneMaxMin$_k$ without careful designs. We prove the same expected runtime of $O(\max\{k,1\}n\ln n)$ for the (G)SEMO, MOEA/D, NSGA-II, and SMS-EMOA. Our brief discussions on a bi-objective LeadingOnes variant with different degrees of conflict show similar findings.
arXiv:2411.16102v2 Announce Type: replace Abstract: Offline batch inference, which leverages the flexibility of request batching to achieve higher throughput and lower costs, is becoming more popular for latency-insensitive applications. Meanwhile, recent progress in model capability and modality makes requests more diverse in compute and memory demands, creating unique opportunities for throughput improvement by resource overlapping. However, a request schedule that maximizes resource overlapping can conflict with the schedule that maximizes prefix sharing, a widely-used performance optimization, causing sub-optimal inference throughput. We present BlendServe, a system that maximizes resource utilization of offline batch inference by combining the benefits of resource overlapping and prefix sharing using a resource-aware prefix tree. BlendServe exploits the relaxed latency requirements in offline batch inference to reorder and overlap requests with varied resource demands while ensuring high prefix sharing. We evaluate BlendServe on a variety of synthetic multi-modal workloads and show that it provides up to $1.44\times$ throughput boost compared to widely-used industry standards, vLLM and SGLang.
arXiv:2501.08238v3 Announce Type: replace Abstract: With the rapid advancement of neural audio codecs, codec-based speech generation (CoSG) systems have become highly powerful. Unfortunately, CoSG also enables the creation of highly realistic deepfake speech, making it easier to mimic an individual's voice and spread misinformation. We refer to this emerging deepfake speech generated by CoSG systems as CodecFake. Detecting such CodecFake is an urgent challenge, yet most existing systems primarily focus on detecting fake speech generated by traditional speech synthesis models. In this paper, we introduce CodecFake+, a large-scale dataset designed to advance CodecFake detection. To our knowledge, CodecFake+ is the largest dataset encompassing the most diverse range of codec architectures. The training set is generated through re-synthesis using 31 publicly available open-source codec models, while the evaluation set includes web-sourced data from 17 advanced CoSG models. We also propose a comprehensive taxonomy that categorizes codecs by their root components: vector quantizer, auxiliary objectives, and decoder types. Our proposed dataset and taxonomy enable detailed analysis at multiple levels to discern the key factors for successful CodecFake detection. At the individual codec level, we validate the effectiveness of using codec re-synthesized speech (CoRS) as training data for large-scale CodecFake detection. At the taxonomy level, we show that detection performance is strongest when the re-synthesis model incorporates disentanglement auxiliary objectives or a frequency-domain decoder. Furthermore, from the perspective of using all the CoRS training data, we show that our proposed taxonomy can be used to select better training data for improving detection performance. Overall, we envision that CodecFake+ will be a valuable resource for both general and fine-grained exploration to develop better anti-spoofing models against CodecFake.
arXiv:2502.01226v4 Announce Type: replace Abstract: Gaussian process (GP) bandits provide a powerful framework for performing blackbox optimization of unknown functions. The characteristics of the unknown function depend heavily on the assumed GP prior. Most work in the literature assume that this prior is known but in practice this seldom holds. Instead, practitioners often rely on maximum likelihood estimation to select the hyperparameters of the prior - which lacks theoretical guarantees. In this work, we study two algorithms for joint prior selection and regret minimization in GP bandits based on GP Thompson sampling (GP-TS): Prior-Elimination GP-TS (PE-GP-TS) that disqualifies priors with poor predictive performance, and HyperPrior GP-TS (HP-GP-TS) that utilizes a bi-level Thompson sampling scheme. We theoretically analyze the algorithms and establish a sublinear regret bound for HP-GP-TS. In addition, we demonstrate the effectiveness of these algorithms compared to the alternatives through extensive experiments with synthetic and real-world data.
arXiv:2502.03616v4 Announce Type: replace Abstract: Noncooperative multi-agent systems often face coordination challenges due to conflicting preferences among agents. In particular, when agents act in their own self-interest, they may prefer different choices among multiple feasible outcomes, leading to suboptimal outcomes or even safety concerns. We propose an algorithm named trading auction for consensus (TACo), a decentralized approach that enables noncooperative agents to reach consensus without communicating directly or disclosing private valuations. TACo facilitates coordination through a structured trading-based auction, where agents iteratively select choices of interest and provably reach an agreement within an a priori bounded number of steps. A series of numerical experiments validate that the termination guarantees of TACo hold in practice, and show that TACo achieves a median performance that minimizes the total cost across all agents, while allocating resources significantly more fairly than baseline approaches.
arXiv:2504.19399v5 Announce Type: replace Abstract: Robust and flexible leader-following is a critical capability for robots to integrate into human society. While existing methods struggle to generalize to leaders of arbitrary form and often fail when the leader temporarily leaves the robot's field of view, this work introduces a unified framework addressing both challenges. First, traditional detection models are replaced with a segmentation model, allowing the leader to be anything. To enhance recognition robustness, a distance frame buffer is implemented that stores leader embeddings at multiple distances, accounting for the unique characteristics of leader-following tasks. Second, a goal-aware adaptation mechanism is designed to govern robot planning states based on the leader's visibility and motion, complemented by a graph-based planner that generates candidate trajectories for each state, ensuring efficient following with obstacle avoidance. Simulations and real-world experiments with a legged robot follower and various leaders (human, ground robot, UAV, legged robot, stop sign) in both indoor and outdoor environments show competitive improvements in follow success rate, reduced visual loss duration, lower collision rate, and decreased leader-follower distance.
arXiv:2505.04753v2 Announce Type: replace Abstract: In this work, we study a six-dimensional movable antenna (6DMA)-enhanced Terahertz (THz) network that supports a large number of users with a few antennas by controlling the three-dimensional (3D) positions and 3D rotations of antenna surfaces/subarrays at the base station (BS). However, the short wavelength of THz signals combined with a large 6DMA movement range extends the near-field region. As a result, a user can be in the far-field region relative to the antennas on one 6DMA surface, while simultaneously residing in the near-field region relative to other 6DMA surfaces. Moreover, 6DMA THz channel estimation suffers from increased computational complexity and pilot overhead due to uneven power distribution across the large number of candidate position-rotation pairs, as well as the limited number of radio frequency (RF) chains in THz bands. To address these issues, we propose an efficient hybrid-field generalized 6DMA THz channel model, which accounts for planar wave propagation within individual 6DMA surfaces and spherical waves among different 6DMA surfaces. Furthermore, we propose a low-overhead channel estimation algorithm that leverages directional sparsity to construct a complete channel map for all potential antenna position-rotation pairs. Numerical results show that the proposed hybrid-field channel model achieves a sum rate close to that of the ground-truth near-field channel model and confirm that the channel estimation method yields accurate results with low complexity.
arXiv:2505.05847v4 Announce Type: replace Abstract: Cuckoo filters are space-efficient approximate set membership data structures with a controllable false positive rate (FPR) and zero false negatives, similar to Bloom filters. In contrast to Bloom filters, Cuckoo filters store multi-bit fingerprints of keys in a hash table using variants of Cuckoo hashing, allowing each fingerprint to be stored at a small number of possible locations. Existing Cuckoo filters use fingerprints of $(k+3)$ bits per key and an additional space overhead factor of at least $1.05$ to achieve an FPR of $2^{-k}$. For $k=10$, this amounts to $1.365\, kn$ bits to store $n$ keys, which is better than $1.443\, kn$ bits for Bloom filters. The $+3$ for the fingerprint size is required to balance out the multiplied FPR caused by looking for the fingerprint at several locations. In the original Cuckoo filter, the number of hash table buckets is restricted to a power of 2, which may lead to much larger space overheads, up to $2.1\, (1+3/k)\, kn$ bits. We present two improvements of Cuckoo filters. First, we remove the restriction that the number of buckets must be a power of 2 by using a different placement strategy. Second, we reduce the space overhead factor of Cuckoo filters to $1.06 \, (1+2/k)$ by using overlapping windows instead of disjoint buckets to maintain the load threshold of the hash table, while reducing the number of alternative slots where any fingerprint may be found. A detailed evaluation demonstrates that the alternative memory layout based on overlapping windows decreases the size of Cuckoo filters not only in theory, but also in practice. A comparison with other state-of-the art filter types, Prefix filters and Vector Quotient filters (VQFs), shows that the reduced space overhead makes windowed Cuckoo filters the smallest filters supporting online insertions, with similarly fast queries, but longer insertion times.
arXiv:2505.21239v2 Announce Type: replace Abstract: Cognitive Diagnosis has become a critical task in AI-empowered education, supporting personalized learning by accurately assessing students' cognitive states. However, traditional cognitive diagnosis models (CDMs) often struggle in cold-start scenarios due to the lack of student-exercise interaction data. Recent NLP-based approaches leveraging pre-trained language models (PLMs) have shown promise by utilizing textual features, but they fail to fully bridge the gap between semantic understanding and cognitive profiling. To address this limitation, we propose \textbf{L}anguage \textbf{M}odel-based \textbf{C}ognitive \textbf{D}iagnosis (LMCD), a unified, LLM-adaptable framework designed to tackle cold-start challenges by harnessing the advanced capabilities of large language models (LLMs). LMCD operates via two primary phases: (1) Knowledge Diffusion, where LLMs generate enriched content for exercises and knowledge concepts (KCs) to establish stronger semantic links; and (2) Semantic-Cognitive Fusion, which leverages LLMs to deeply integrate textual information with student cognitive states. By unifying the semantic and cognitive spaces, LMCD creates comprehensive representations that serve as a plug-and-play enhancement for various off-the-shelf CDMs. Experiments on two real-world datasets demonstrate that LMCD significantly outperforms state-of-the-art methods in both exercise-cold and domain-cold settings. https://github.com/TAL-auroraX/LMCDThe code is publicly available at https://github.com/TAL-auroraX/LMCD
arXiv:2507.07805v3 Announce Type: replace Abstract: Safety filters for industrial constrained systems are required to combine certified constraint satisfaction, predictable online computation, and a transparent tuning interface. Existing set-based filters are based on a well-established control invariant set design that scales favorably with state and input constraints, but typically intervene only at the set boundary. Control barrier function (CBF)-based filters, by contrast, provide tunable intervention but require a scalar barrier construction. This paper proposes a set-based CBF safety filter that turns a convex control invariant set directly into a tunable barrier via its Minkowski functional. The resulting filter is formulated as a single-level quadratic program (QP) in which one class-$\mathcal{K}^e$ parameter sets the intervention aggressiveness. Explicit convex formulations are derived for polytopic, zonotopic, and MPC-based invariant sets. Under standard bounded-disturbance assumptions, the resulting safety filter guarantees constraint satisfaction and asymptotic recovery into the invariant set. For tight real-time budgets, a learning-based approximation enables online acceleration, while the formal safety guarantees remain tied to the exact formulation. The method is validated in numerical studies and on a permanent-magnet synchronous motor drive, where an explicit QP implementation evaluates within a 150 microseconds sampling window and has a worst-case execution time of 28.04 microseconds.
arXiv:2507.09092v2 Announce Type: replace Abstract: With the intervention of machine vision in our crucial day to day necessities including healthcare and automated power plants, attention has been drawn to the internal mechanisms of convolutional neural networks, and the reason why the network provides specific inferences. This paper proposes a novel post-hoc visual explanation method called MI CAM based on activation mapping. Differing from previous class activation mapping based approaches, MI CAM produces saliency visualizations by weighing each feature map through its mutual information with the input image and the final result is generated by a linear combination of weights and activation maps. It also adheres to producing causal interpretations as validated with the help of counterfactual analysis. We aim to exhibit the visual performance and unbiased justifications for the model inferencing procedure achieved by MI CAM. Our approach works at par with all state-of-the-art methods but particularly outperforms some in terms of qualitative and quantitative measures.
arXiv:2509.01739v3 Announce Type: replace Abstract: Multi-objective evolutionary algorithms (MOEAs) are among the most widely and successfully applied optimizers for multi-objective problems. However, to store many optimal trade-offs (the Pareto optima) simultaneously, MOEAs are typically run with a large population of solution candidates. This slows down the algorithm and renders the choice of the population size a crucial design decision. In this work, we aim to overcome these difficulties by proposing the dynamic NSGA-II, a variant of the well-known NSGA-II that starts with a small initial population and doubles it after a user-specified number $\tau$ of function evaluations, up to a maximum size of $N_{max}$. We prove that the dynamic NSGA-II with optimal parameters computes the Pareto front of the OneMinMax benchmark of size $n$ with high probability in $O(n \log^2 n)$ function evaluations, which is considerably faster than the $\Theta(n^2 \log n)$ runtime of the static NSGA-II with optimal parameters. For the OneJumpZeroJump benchmark with gap size $k$, we show a runtime of $O(n^k \log^2 n)$, improving upon the known runtime of $\Theta(n^{k+1})$. We also propose a variant that uses the initial population size for a longer period and achieves slightly better performance. Finally, we show that a simple concurrent-run strategy turns our dynamic NSGA-II variants into parameter-less algorithms that exceed the above runtimes only by a logarithmic factor and hence still outperform the static NSGA-II by a factor of $\tilde\Omega(n)$.
arXiv:2509.09151v2 Announce Type: replace Abstract: Research in video understanding has advanced rapidly, driven by increasingly diverse datasets and more powerful model architectures. While existing surveys typically organize progress by tasks, benchmarks, or model families, they provide limited insight into why particular architectures emerged and succeeded. In this survey, we argue that the evolution of video understanding is fundamentally shaped by dataset structure. We present a dataset-centric perspective that connects dataset structure, inductive biases, and architectural design within a unified framework. We show that different datasets require models to capture specific invariances and capabilities, such as robustness to viewpoint changes, sensitivity to temporal ordering, reasoning over long-range dependencies, relational interactions, and cross-modal alignment. These requirements naturally give rise to inductive biases, i.e., architectural assumptions that favor particular patterns of reasoning and generalization. From this perspective, milestone architectures, including two-stream networks, 3D CNNs, temporal models, transformers, graph-based methods, and multimodal foundation models, can be understood as architectural responses to the challenges posed by evolving datasets. Building on this framework, we systematically analyze how dataset characteristics have shaped architectural innovation across video understanding tasks and discuss the representational biases induced by different data regimes. By unifying datasets, inductive biases, and architectures into a coherent perspective, this survey offers both a retrospective explanation of the field's evolution and a forward-looking roadmap toward general-purpose video understanding systems. Code and dynamic video visualizations of dataset-induced biases are available at https://time.griffith.edu.au/paper-sites/video-understanding/.
arXiv:2509.22097v5 Announce Type: replace Abstract: Large language model-powered code agents are rapidly transforming software engineering, yet the security risks of their generated code have become a critical concern. Existing benchmarks have provided valuable insights, but they fail to capture scenarios in which vulnerabilities are actually introduced by human developers, making fair comparisons between humans and agents infeasible. We therefore introduce SecureVibeBench, a benchmark of 105 C/C++ secure coding tasks sourced from 41 projects in OSS-Fuzz for code agents. SecureVibeBench has the following features: (i) realistic task settings that require multi-file edits in large repositories, (ii)~aligned contexts based on real-world open-source vulnerabilities with precisely identified vulnerability introduction points, and (iii) comprehensive evaluation that combines functionality testing and security checking with both static and dynamic oracles. We evaluate 5 popular code agents like OpenHands, supported by 5 LLMs (e.g., Claude sonnet 4.5) on SecureVibeBench. Results show that current agents struggle to produce both correct and secure code, as even the best-performing one, produces merely 23.8\% correct and secure solutions on SecureVibeBench. Our code and data are on https://github.com/iCSawyer/SecureVibeBench.
arXiv:2510.21290v3 Announce Type: replace Abstract: Partial Differential Equations (PDEs) are fundamental mathematical models for describing physical phenomena, yet most PDEs of practical interest require numerical approximations. The feasibility of such methods is constrained by existing computational models. Since digital computers are the primary realizations of numerical computations, and Turing machines define their theoretical limits, computability of PDE solutions is of fundamental significance. It provides a rigorous framework to distinguish equations that are effectively solvable from those that encode undecidable or non-computable behavior. Once computability is established, complexity theory quantifies the resources required to approximate PDE solutions. In this work, we present a novel framework based on least-squares variational formulations and associated gradient flows to analyze the computability and complexity of PDE solutions from an optimization perspective. Our approach approximates PDE solution operators via discrete gradient flows, linking PDE properties, such as coercivity, ellipticity, and convexity, to solution complexity. Within this setting, we characterize representation- and discretization-dependent sufficient conditions for regimes where PDEs admit polynomial-time approximations, as well as regimes exhibiting complexity blowup, where polynomial-time input data produce solutions with super-polynomial complexity. In summary, this paper develops a variational framework for analyzing computability and computational complexity of PDE solution classes. The results show how PDE structure and solution regularity influence their complexity, by establishing sufficient conditions for computability and complexity bounds. Beyond the theoretical characterization, the framework provides guidelines for effective numerical methods and contributes to understanding the limitations of digital computation for PDE problems.
arXiv:2510.22024v2 Announce Type: replace Abstract: Modern connected vehicles rely on persistent LTE connectivity to enable remote diagnostics, over-the-air (OTA) updates, and safety-relevant services. While mobile network vulnerabilities are well documented in the smartphone ecosystem, their impact in safety-relevant automotive settings remains insufficiently examined. We conduct a black-box case study of LTE security in Tesla's Model 3 and Cybertruck, revealing systemic protocol weaknesses and architectural misconfigurations in connected vehicles. We find that Tesla's telematics stack is susceptible to IMSI catching, rogue base station hijacking, and insecure fallback mechanisms that may silently degrade service availability. Furthermore, legacy control-plane configurations allow for silent SMS injection and broadcast message spoofing without driver awareness. While the vulnerabilities are grounded in Tesla, this case study suggests broader implications for connected-vehicle telematics and for regulatory frameworks such as ISO/SAE 21434 and UN R155/R156, which assume secure, traceable, and resilient telematics in modern vehicles.
arXiv:2511.05355v3 Announce Type: replace Abstract: Flow matching (FM) has shown promising results in data-driven planning. However, it inherently lacks formal guarantees for ensuring state and action constraints, whose satisfaction is a fundamental and crucial requirement for the safety and admissibility of planned trajectories on various systems. Moreover, existing FM planners do not ensure the dynamical consistency, which potentially renders trajectories inexecutable. We address these shortcomings by proposing SAD-Flower, a novel framework for generating Safe, Admissible, and Dynamically consistent trajectories. Our approach relies on an augmentation of the flow with a virtual control input. Thereby, principled guidance can be derived using techniques from nonlinear control theory, providing formal guarantees for state constraints, action constraints, and dynamic consistency. Crucially, SAD-Flower operates without retraining, enabling test-time satisfaction of unseen constraints. Through extensive experiments across several tasks, we demonstrate that SAD-Flower outperforms various generative-model-based baselines in ensuring constraint satisfaction.
arXiv:2511.07002v2 Announce Type: replace Abstract: Even though we know the precise computations that lead from a large language model (LLM) input to its output this computation remains very hard to interpret. One way to make it easier to understand this process is by creating a sparse computational graph that captures most of the model behavior with smallest number of computational nodes. Cross-layer transcoders (CLT) decompose the dense computations of the MLP but the resulting circuits still contain thousands of nodes even for short prompts. Existing automated interpretation methods label individual features from corpus activations, and it often happens that these labels are not validated by causal intervention. We introduce probe prompting, a transparent rule-based pipeline that groups the features of an attribution graph into concept-aligned supernodes from their responses on a small set of concept-targeted probe prompts, summarized as Cross-Prompt Activation Signatures (CPAS). Across four factual domains, on Gemma-2-2B with a public CLT dictionary and 45,596 entity-swap interventions, we find that the labeled supernodes have the predicted steering behavior in every one of them. Code, datasets, and an interactive demo are released anonymously as a reusable harness for calibrating supernode labels against causal interventions.
arXiv:2511.07317v2 Announce Type: replace Abstract: We introduce Reinforcement Learning (RL) with Adaptive Verifiable Environments (RLVE), an approach using verifiable environments that procedurally generate problems and provide algorithmically verifiable rewards, to scale up RL for language models (LMs). RLVE enables each verifiable environment to dynamically adapt its problem difficulty distribution to the policy model's capabilities as training progresses. In contrast, static data distributions often lead to vanishing learning signals when problems are either too easy or too hard for the policy. To implement RLVE, we create RLVE-Gym, a large-scale suite of 400 verifiable environments carefully developed through manual environment engineering. Using RLVE-Gym, we show that environment scaling, i.e., expanding the collection of training environments, consistently improves generalizable reasoning capabilities. RLVE with joint training across all 400 environments in RLVE-Gym yields a 3.37% absolute average improvement across six reasoning benchmarks, starting from one of the strongest 1.5B reasoning LMs. By comparison, continuing this LM's original RL training yields only a 0.49% average absolute gain despite using over 3x more compute. We release our code publicly.
arXiv:2511.07938v3 Announce Type: replace Abstract: Power-logistics scheduling in modern seaports typically follows a predict-then-optimize pipeline. To enhance the decision quality of predictions, decision-focused learning has been proposed, which aligns the training of forecasting models with downstream decision outcomes. However, this end-to-end design inherently restricts the value of forecasting models to a specific task structure and therefore generalizes poorly to evolving tasks induced by varying vessel arrivals. We address this gap with a decision-focused continual learning framework that adapts online to a stream of scheduling tasks. Specifically, we introduce Fisher-information-based regularization to enhance cross-task generalization by preserving parameters critical to prior tasks. A differentiable convex surrogate is also developed to stabilize gradient backpropagation. The proposed approach enables learning a decision-aligned forecasting model across a varying task stream with sustainable long-term computational and memory requirements. Experiments calibrated to Jurong Port show improved decision performance and cross-task generalization over existing methods, together with reduced computational cost and a bounded memory footprint.
arXiv:2511.11012v2 Announce Type: replace Abstract: Automated program repair has traditionally focused on single-hunk defects, overlooking multi-hunk bugs that are prevalent in real-world systems. Repairing these bugs requires coordinated edits across multiple, disjoint code regions, posing substantially greater challenges. We present the first systematic study of LLM-driven coding agents (Claude Code, Codex, Gemini-cli, and Qwen Code) on this task. We evaluate these four state-of-the-art agents on 404 multi-hunk bugs from the PolyHunk dataset, yielding 1,616 repair trajectories for large-scale behavioral analysis. We employ fine-grained metrics to assess localization, repair accuracy, regression behavior, and operational dynamics across agents. We find that localization capability varies substantially, with Codex achieving the highest success rate (75.3%) and Qwen Code the lowest (40.4%). Repair accuracy also differs widely, ranging from 26.98% (Qwen Code) to 92.82% (Claude Code), and consistently declines with increasing bug dispersion and complexity (hunk divergence and spatial proximity). High-performing agents (Claude Code and Codex) demonstrate superior semantic consistency, achieving positive average regression reduction, whereas lower-performing agents often introduce new test failures. Notably, agents do not fail fast; failed repairs consume substantially more resources (33%-440% more input tokens) and require longer execution time (35%-330%). Additionally, we developed Maple to provide agents with repository-level context. Empirical results show that Maple improves repair accuracy of Gemini-cli by ~21% through enhanced localization. By analyzing fine-grained metrics and trajectory-level analysis, this study moves beyond accuracy to explain how coding agents localize, reason, and act during multi-hunk repair. Our findings underscore the impact of bug divergence and spatial proximity on multi-hunk repair success for coding agents.
arXiv:2511.18421v2 Announce Type: replace Abstract: Existing Test-time Adaptation (TTA) studies rely heavily on static and homogeneous corruption protocols, such as ImageNet-C and CIFAR-10-C/100-C, leading to inconsistent evaluation settings and potentially inflated robustness estimates that are compared with real-world situations. TTA lacks a standardized evaluation infrastructure capable of modeling realistic heterogeneous acoustic degradation. We introduce DHAuDS, a standardized benchmark suite for evaluating audio classification TTA robustness under dynamic corruption severity and heterogeneous noise mixtures. Rather than proposing a new TTA algorithm, DHAuDS focuses on exposing robustness limitations that remain hidden under conventional fixed-noise evaluation protocols.
arXiv:2511.19829v3 Announce Type: replace Abstract: Prompt optimization has become a central mechanism for eliciting strong performance from LLMs, and recent work has made substantial progress by proposing diverse prompt evaluation metrics and optimization strategies. Despite these advances, prompt evaluation and prompt optimization are often developed in isolation, limiting the extent to which evaluation can effectively inform prompt refinement. In this work, we study prompt optimization as a process guided by performance-relevant evaluation signals. To address the disconnect between evaluation and optimization, we propose an evaluation-instructed prompt optimization approach that explicitly connects prompt evaluation with query-dependent optimization. Our method integrates multiple complementary prompt quality metrics into a performance-reflective evaluation framework and trains an execution-free evaluator that predicts prompt quality directly from text, avoiding repeated model executions. These evaluation signals then guide prompt refinement in a targeted and interpretable manner. Empirically, the proposed evaluator achieves 83.7% accuracy in predicting prompt performance. When incorporated into the optimization process, our approach consistently outperforms existing optimization baselines across eight benchmark datasets and three different backbone LLMs. Overall, our results demonstrate that reliable and efficient evaluation signals can serve as an effective foundation for robust and interpretable prompt optimization.