arXiv:2607.05058v1 Announce Type: new Abstract: Technical lectures mix Mandarin speech with English technical terms. These terms carry the core meaning of the lecture, yet they occupy few characters. Character error rate (CER) therefore hides their recognition failures. We study whether lecture context helps recognize these terms. We build a term-rich Mandarin AI/ML lecture benchmark, and we define term-centric metrics that measure technical-term recognition directly. We then propose a two-pass, reference-free decoding method. The first pass runs segment-only ASR. We extract the most frequent technical terms from the first-pass hypotheses, and we prompt the recognizer with this self-built glossary in the second pass. Across five ASR backbones, the first-pass glossary raises term recall for every model and holds or lowers CER on all five. On Breeze-ASR-25 it lifts term recall from 52.50% to 60.13% while lowering CER, and a hybrid that adds a small external term list reaches 62.05% recall and 82.73% term precision. Lecture context, recovered from the model's own output, is a practical signal for technical-term recognition. Term-centric evaluation exposes errors that CER misses.
Science Journals
arXiv:2507.04032v3 Announce Type: replace Abstract: We introduce remarkable upper bounds, which are sharp and given by simple formulas, for the interpolation error constants on triangles. These constants are crucial for analyzing interpolation errors, particularly those associated with the finite element method. We prove boundedness via a numerical verification method and asymptotic analysis. The proof process used here can be applied to various other norm inequalities.
arXiv:2601.10041v2 Announce Type: replace Abstract: Emergency department (ED) overcrowding and patient boarding represent critical systemic challenges that compromise care quality. We propose a threshold-based admission policy that redirects non-urgent patients to alternative care pathways, such as telemedicine, during peak congestion. The ED is modeled as a two-class $M/M/c$ preemptive-priority queuing system, where high-acuity patients are prioritized and low-acuity patients are subject to state-dependent redirection. Analyzed via a level-dependent Quasi-Birth-Death (QBD) process, the model determines the optimal threshold by maximizing a long-run time-averaged objective function comprising redirection-affected revenue and costs associated with patient balking and system occupancy. Structural analysis establishes monotone comparative statics relating the optimal threshold to all model parameters analytically. Numerical analysis using national healthcare data reveals that optimal policies are highly context-dependent. While rural EDs generally optimize at lower redirection thresholds, urban EDs exhibit performance peaks at moderate thresholds. Results indicate that our optimal policy yields significant performance gains of up to $4.84\%$ in rural settings and $5.90\%$ in urban environments. This research provides a mathematically rigorous framework for balancing clinical priority with operational efficiency across diverse ED settings.
arXiv:2604.03844v3 Announce Type: replace Abstract: Tokenized assets increasingly operate across heterogeneous blockchain networks and off-chain ledgers, where a regulatory action (a freeze, a seizure, a confiscation) must take effect atomically and consistently across every domain holding the asset. We mechanize, in Isabelle/HOL, cross-domain state preservation as a functor: state machines are objects, structure-preserving synchronization maps are morphisms, and the category laws (identity, composition, associativity) hold as theorems. On this base we establish four results. Safety: a regulatory transition on one domain is faithfully reflected across all connected domains, with bidirectional roundtrip preservation, N-domain consistency, per-asset isolation, and terminal states preserved. Liveness: under f < n/3 Byzantine nodes, deterministic conflict resolution and starvation freedom under a fair-leader assumption, with the threshold n >= 3f+1 shown to make that assumption inhabitable rather than vacuous. Convergence: from an arbitrary unlocked configuration, with no initial cross-chain consistency assumed, synchronization reaches a valid state in a bounded number of steps along a terminal-faithful recovery path. Hierarchy: a tower of synchronization-degree functors connected by natural transformations closed under composition, with a genuinely one-directional degree monotonicity. We couple the functor to Lochbihler and Maric's authenticated data structure at the global-state level, instantiated on a recursive model of the Canton transaction tree with a declared consensus-scope limit. The synchronization model is atomic; its lift to a partially synchronous network is future work. The application is a regulatory state transition model distilled from the RCP framework (arXiv:2603.29278). All ten Isabelle/HOL theory files build without sorry or oops and are submitted to the Archive of Formal Proofs.
arXiv:2411.11449v3 Announce Type: replace Abstract: Deploying AI systems in public institutions can have far-reaching consequences for many people, making it a matter of public interest. Providing opportunities for stakeholders to come together, understand these systems, and debate their merits and harms is thus essential. Explainable AI often focuses on individuals, but deliberation benefits from group settings, which are underexplored. To address this gap, we present findings from an interview study with 8 focus groups and 12 individuals. Our findings provide insight into how explanations support AI novices in deliberating alone and in groups. Participants used modular explanations with four information categories to solve tasks and decide about an AI system's deployment. We found that the explanations supported groups in creating shared understanding and in finding arguments for and against the system's deployment. In comparison, individual participants engaged with explanations in more depth and performed better in the study tasks, but missed an exchange with others. Based on our findings, we provide suggestions on how explanations should be designed to work in group settings and describe their potential use in real-world contexts. With this, our contributions inform XAI research that aims to enable AI novices to understand and deliberate AI systems in the public sector.
arXiv:2412.00393v3 Announce Type: replace Abstract: Analyzing process data at varying levels of granularity is important to derive actionable insights and support informed decision-making. Object-Centric Event Data (OCED) enhances process mining by capturing interactions among events and multiple objects, leading to the discovery of more detailed and realistic yet complex process models. The lack of methods to adjust the granularity of the analysis prevents users from leveraging the full potential of Object-Centric Process Mining (OCPM). To address this gap, we propose four operations: drill-down, roll-up, unfold, and fold, which enable analysts to change the granularity of analysis when working with Object-Centric Event Logs (OCEL). These operations allow analysts to seamlessly transition between detailed and aggregated process models, facilitating the discovery of insights that require varying levels of abstraction. We formally define these operations and implement them in an open-source Python library. To validate their utility, we applied the approach to real-world OCEL data extracted from a learning management system, covering a four-year period and approximately 400 students, as a case of object-centric educational process mining. This case study shows significant improvements in the precision and fitness of the discovered models after applying the operations. In addition, we evaluate the scalability of the operators on large, publicly available OCELs derived from the Business Process Intelligence Challenge datasets, demonstrating that the operations remain computationally feasible on industrial-scale event logs. This approach can empower analysts to perform more flexible and comprehensive process exploration, unlocking actionable insights through flexible granularity adjustments.
arXiv:2604.08061v2 Announce Type: replace Abstract: Hydrogen flames exhibit multiple intrinsic instabilities. The low molar masses of H and H2 lead to significant Soret diffusion near the flame front; however, its influence on hydrogen flame instabilities remains to be quantified. This study investigates the effect of Soret diffusion on instability evolution dynamics via one-dimensional counterflow analysis and two-dimensional, high-fidelity direct numerical simulations covering both the linear growth regime and the fully developed nonlinear regime over a wide range of equivalence ratios (phi). In the linear regime, Soret diffusion increases the perturbation growth rate at phi < 1.7, especially under lean conditions, but reduces the growth rate at phi > 1.7. A similar sensitivity reversal is observed in the Markstein length near the peak equivalence ratio of unstretched laminar flame speed. In the nonlinear regime, Soret diffusion accelerates the formation of small-scale wrinkles in lean hydrogen flames and reduces the characteristic size of large-scale finger structure by one-third. An interesting observation is that, although Soret diffusion promotes preferential diffusion and increases the local flame displacement speed, the global fuel consumption rate decreases due to a reduction in the overall flame surface area. In addition, curvature-based flame segment analysis reveals a synergistic effect between Soret diffusion and Fickian diffusion that enhances/reduces the local equivalence ratio in positively/negatively curved regions of the flame front. The probability distributions of the Karlovitz number and the density-weighted displacement speed are also analyzed; results suggest that, for lean hydrogen flames, Soret diffusion broadens the distributions for both parameters, particularly on the positive side. These findings promise to advance the fundamental understanding of hydrogen flame dynamics under complex differential transport.
arXiv:2604.10062v3 Announce Type: replace Abstract: We study reward poisoning attacks in reinforcement learning (RL), where an adversary manipulates rewards within constrained budgets to force the target RL agent to adopt a policy that aligns with the attacker's objectives. Prior works on reward poisoning mainly focused on sufficient conditions to design a successful attacker, while only a few studies discussed the infeasibility of targeted attacks. This paper provides the first precise necessity and sufficiency characterization of the attackability of a linear MDP under reward poisoning attacks. Our characterization draws a bright line between the vulnerable RL instances, and the intrinsically robust ones which cannot be attacked without large costs even running vanilla non-robust RL algorithms. Our theory extends beyond linear MDPs -- by approximating deep RL environments as linear MDPs, we show that our theoretical framework effectively distinguishes the attackability and efficiently attacks the vulnerable ones, demonstrating both the theoretical and practical significance of our characterization.
arXiv:2604.10311v2 Announce Type: replace Abstract: Artificial Intelligence (AI) models, encompassing both traditional machine learning (ML) and more advanced approaches such as deep learning and large language models (LLMs), play a central role in modern applications. AI model lifecycle management involves the end-to-end process of managing these models, from data collection and preparation to model building, evaluation, deployment, and continuous monitoring. This process is inherently complex, as it requires the coordination of diverse services that manage AI artifacts such as datasets, dataflows, and models, all orchestrated to operate seamlessly. In this context, it is essential to isolate applications from the complexity of interacting with heterogeneous services, datasets, and AI platforms. In this paper, we introduce Gypscie, a cross-platform AI artifact management system. By providing a unified view of all AI artifacts, the Gypscie platform simplifies the development and deployment of AI applications. This unified view is realized through a knowledge graph that captures application semantics and a rule-based query language that supports reasoning over data and models. Model lifecycle activities are represented as high-level dataflows that can be scheduled across multiple platforms, such as servers, cloud platforms, or supercomputers. Finally, Gypscie records provenance information about the artifacts it produces, thereby enabling explainability. Our qualitative comparison with representative AI systems shows that Gypscie supports a broader range of functionalities across the AI artifact lifecycle. Our experimental evaluation demonstrates that Gypscie can successfully optimize and schedule dataflows on AI platforms from an abstract specification.
arXiv:2604.11062v2 Announce Type: replace Abstract: This paper proposes a coding framework for capacity-region-achieving sparse regression (SR) codes over MIMO multiple-access channels (MIMO-MAC), where a single SR code is used for each user at the transmitter. With random semi-unitary dictionary matrices applied for encoding, multiple-access OAMP (MA-OAMP) enables reliable parallel interference cancellation (PIC) at the receiver. Theoretically, an optimal coding principle with the MA-OAMP receiver, which achieves the sum capacity and, in combination with time sharing, achieves the entire capacity region, is established as the guiding principle for designing capacity-region-achieving codes. Accordingly, a coding scheme for capacity-region-achieving SR codes is proposed via proper power allocation over the position-modulated signals.
Multimodal Ambivalence/Hesitancy Recognition in Videos for Personalized Digital Health Interventions
arXiv:2604.11730v4 Announce Type: replace Abstract: Using behavioural science, health interventions focus on behaviour change by providing a framework to help patients acquire and maintain healthy habits that improve medical outcomes. In-person interventions are costly and difficult to scale, especially in resource-limited regions. Digital health interventions offer a cost-effective approach, potentially supporting independent living and self-management. Automating such interventions, especially through machine learning, has recently gained considerable attention. Ambivalence and hesitancy (A/H) play a primary role for individuals to delay, avoid, or abandon health interventions. A/H are subtle and conflicting emotions that place a person in a state between positive and negative evaluations of a behaviour, or between acceptance and refusal to engage in it. They manifest as affective inconsistency across modalities or within a modality, such as language, facial, vocal expressions, and body language. While experts can be trained to recognize A/H, integrating them into digital health interventions is costly and less effective. Automatic A/H recognition is therefore critical for the personalization and cost-effectiveness of digital health interventions. Here, we explore the application of deep learning models for A/H recognition in videos, a multi-modal task by nature. In particular, this paper covers three learning setups: supervised learning, unsupervised domain adaptation for personalization, and zero-shot inference via large language models (LLMs). Our experiments are conducted on the unique and recently published BAH video dataset for A/H recognition. Our results show limited performance, suggesting that more adapted multi-modal models are required for accurate A/H recognition. Better methods for modeling spatio-temporal and multimodal fusion are necessary to leverage conflicts within/across modalities.
arXiv:2504.20412v3 Announce Type: replace Abstract: Fuzzing frameworks like syzkaller have uncovered thousands of Linux kernel crashes, many of which are critical and security-sensitive. However, the ability to rapidly repair these crashes has not kept pace, particularly given the complexity and low-level nature of kernel code. Predominantly targeting user-space applications, existing LLM-based program repair techniques are not tailored to the unique challenges posed by kernel fuzz bugs-such as the absence of natural language bug reports, lack of exhaustive test oracles, and highly specialized crash artifacts. Thus, in this work, we first identify the prevalent bottlenecks that generic agents struggle with in complex systems such as the Linux kernel. Guided by these findings, we then build kAgent, a workflow-based agent inspired by how kernel developers diagnose and fix bugs, and kGym++, the co-designed toolstack supporting kAgent's requests. kAgent inspects relevant execution logs, generates execution-grounded natural language hypotheses, synthesizes candidate patches, validates patches through crash reproduction, and iteratively refines its reasoning. We ablate these agentic system features in kAgent and quantitatively analyze their contributions to the overall performance. We also report our experience on building agents for kernel crash repair. Although derived from this work on Linux, we note that our experience applies broadly to fuzzing-discovered bugs in complex systems software. We evaluate kAgent on kBenchSyz and show that it can repair up to 54.5% of crashes without localization and 65% with correct file hints. We also show kAgent's generalization on a few wild Syzkaller bugs, and how different patch types offer varying utility to developers who debug complex system software.
arXiv:2607.05005v1 Announce Type: new Abstract: Low-light degradation reduces image visibility and weakens structural cues that are important for visual representation and scene understanding. Existing low-light image enhancement methods mainly focus on appearance restoration, while insufficiently exploiting scene geometry to preserve structural consistency. To address this limitation, this paper proposes a Depth-guided Multi-scale Attention Network (DMSA-Net) for geometry-aware low-light image enhancement. DMSA-Net introduces depth-related structural priors into low-light representation learning through reflectance-geometry interaction. A Retinex-based decomposition module is first used to obtain illumination-invariant reflectance representations, from which depth cues are inferred to characterize scene structure under degraded illumination. A multi-scale depth-guided fusion strategy is then embedded into a hierarchical encoder-decoder architecture, where depth-aware attention adaptively integrates geometric and appearance features. Experiments on several benchmark datasets show that DMSA-Net achieves effective low-light restoration while improving structural preservation. Moreover, we construct LOL-D, a depth-augmented low-light dataset, to facilitate research on geometry-aware low-light vision.
arXiv:2607.05123v1 Announce Type: new Abstract: Recent advances in large language models and programmatic CAD have significantly improved Text-to-CAD generation for individual parts. However, production-ready mechanical assembly generation remains largely unsolved. Unlike single-part modeling, assemblies require coordinated reasoning over multiple components, functional interfaces, assembly relations, engineering principles, and physical consistency. Consequently, directly generating executable CAD code is insufficient for constructing mechanically valid and reusable assemblies. We present AssemCAD, an axiom-grounded framework for production-ready CAD assembly generation from natural language. Instead of representing an assembly as monolithic CAD code, AssemCAD first constructs an axiomatic Assembly Specification consisting of typed parts, geometry-backed ports, executable mates, and engineering axioms. Each assembly relation is explicitly grounded in one or more engineering principles, making the resulting specification interpretable, reusable, and verifiable. To realize this specification, AssemCAD introduces a port- and mate-based CAD assembly library that executes symbolic assembly relations through deterministic mate transformations and validates declared interfaces using concrete B-Rep geometric evidence. Built on this representation and library, AssemCAD further supports on-demand synthesis of reusable parametric component factories for both standard and open-world geometries. Experiments on AssemBench show that AssemCAD substantially improves assembly preservation and physical validity over code-centric CAD generation baselines, while generalizing across different foundation-model backbones. By combining axiom-grounded assembly reasoning with deterministic geometric execution, AssemCAD extends Text-to-CAD from isolated part generation toward production-ready mechanical assembly design.
arXiv:2607.05008v1 Announce Type: new Abstract: Echocardiography is the first imaging modality used for assessing cardiac function, and accurate segmentation of cardiac structures is essential for deriving biomarkers. However, the development of effective automated segmentation models for multiple cardiac structures is challenged by the difficulty of training on datasets from different sources that are often partially-labelled. This study aims to address this challenge by evaluating the performance of three loss functions - adaptive categorical cross entropy (aCCE) loss, marginal loss, and the adaptive binary cross entropy (aBCE) loss - in handling partially-labelled data. We conduct a comprehensive comparison of these loss functions across multiple scenarios and network architectures: intra-domain and inter-domain tasks, with both single and multiple partial-labels, and varying proportions of fully-labelled to partially-labelled data. Our experiments reveal that all three loss functions exhibit strong performance in intra-domain segmentation tasks, effectively handling label variations within the same domain. For inter-domain tasks, where models are trained on datasets with a domain shift, the aBCE and marginal losses show superior performance when dealing with the case of one label being missing from some training examples. In scenarios involving more than one label being missing, marginal loss outperforms the other methods, demonstrating its robustness in such complex conditions. These results highlight the strengths of each loss function depending on the labelling scenario, emphasizing the importance of selecting the appropriate loss function to optimize model performance. This study represents the first investigation of techniques for handling partially-labelled data from multiple different domains in echocardiography segmentation and provides a comprehensive comparison of loss-based solutions.
arXiv:2607.05009v1 Announce Type: new Abstract: Complete digital 12-lead electrocardiograms (ECGs) are essential for AI-enabled cardiovascular assessment, yet many clinical ECG records, particularly those digitized from ECG images, remain incomplete because of short display formats, incomplete waveform digitization, lead loss, or signal corruption. We developed ImputeECG, a mask-conditioned one-dimensional Transformer autoencoder that completes 12-lead, 10-s ECGs while retaining all observed samples. The model was trained on PTB-XL and evaluated on PTB-XL and CPSC2018 under simulated incomplete settings, with additional real-world validation in a 43,633-record Kailuan clinical cohort after ECG image digitization. Metrics were computed over originally missing regions, with analyses of morphology and downstream diagnostic utility. On PTB-XL, ImputeECG reduced missing-region MAE by 41.7-51.0% and MSE by 54.0-63.7% versus the strongest baseline, with lower errors in R-peak timing, RR interval, QRS duration, QT interval, and P-wave, QRS-complex, and T-wave reconstruction. On CPSC2018, ImputeECG reduced MAE by 49.7-51.9%, supporting external generalization. In downstream multi-label classification, ImputeECG restored performance to 92.28% AUROC and 33.88% AUPRC in the most incomplete PTB-XL setting, approaching complete-ECG performance. On CPSC2018, completed ECGs achieved 94.75-95.89% AUROC and 78.83-81.86% AUPRC across settings. In Kailuan, ECG completion improved zero-shot sex prediction AUROC from 82.6% to 85.8% and reduced age prediction MAE from 10.72 to 9.87 years after image-based ECG digitization. These findings support ECG completion as a practical strategy for converting incomplete ECG records into AI-ready 12-lead, 10-s digital signals and extending the usable scope of ECG archives for digital cardiac assessment.
arXiv:2607.04749v1 Announce Type: new Abstract: Large language model (LLM) training shall adapt to dynamic resources in shared clusters to tackle the elasticity, including passive preemption and optimistic scaling. State migration across device sets is required when altering the hybrid-parallel configuration due to dynamic resources. Existing solutions rely on checkpoint-based mechanisms, which persist complete states to storage for resuming with re-assigned resources, forcing all GPUs to stall when transferring model states. Despite optimization efforts, checkpoint-based solutions incur prohibitive latency due to data movement across memory hierarchies. We propose ETC, a checkpoint-free state migration framework for elastic hybrid-parallel LLM training. We exploits the state locality to minimize inter-GPU data movement, replacing storage persistence with direct peer-to-peer communication. Besides, we eliminate node fragmentation through communication coalescing. Integrated with Megatron-LM, ETC reduces migration overhead by 2.33$\times$ to 6.37$\times$ compared to checkpoint-based solutions across diverse parallel configurations. By enabling efficient migration, ETC unlocks practical elastic training in production environments.
arXiv:2607.05013v1 Announce Type: new Abstract: Although LLMs have made significant progress in mathematical reasoning, determining whether a mathematical problem is solvable remains a fundamental yet challenging capability. While recent studies have probed internal representations of model solvability beliefs, verbalization has primarily been studied behaviorally rather than as an internal representation, limiting its analysis and manipulation. We address this gap by separately probing representations of solvability knowledge and verbalization, allowing us to disentangle the two within model hidden states. Across multiple LLMs, we show that knowledge and verbalization are encoded as distinct, linearly decodable representations and that fabrication is primarily associated with changes in verbalization rather than the underlying knowledge. Prompting with unsolvability cues reduces fabrication primarily by shifting verbalization, while activation steering demonstrates that these representations can be echanistically manipulated to improve model abstention.
arXiv:2604.11770v2 Announce Type: replace Abstract: Automated Program Repair (APR) has recently benefited from large language models (LLMs). However, most LLM-based APR approaches still rely primarily on coarse-grained, end-to-end signals from test-suite outcomes or specification checks on the candidate fix to guide repair, providing limited insight into where a program's internal logic deviates from its intended behavior. In contrast, human debugging often relies on intermediate reasoning about program states through localized correctness conditions or assertions. Inspired by this observation, we propose SpecTune, a specification-guided debugging framework that incorporates intermediate behavioral reasoning into APR. SpecTune decomposes the repair task into suspicious regions connected by execution checkpoints and derives localized postconditions representing expected program behaviors at those points. By executing the buggy program and evaluating these postconditions, SpecTune produces micro-level debugging signals that indicate mismatches between observed and intended behaviors, enabling more precise fault localization and targeted patch generation. To address the potential unreliability of LLM-generated postconditions, we introduce two complementary signals: a specification validation signal alpha, which estimates the consistency of generated postconditions using partial passing test cases, and a discriminative signal beta, which detects violations of validated postconditions during execution. With these signals, SpecTune safely leverages automatically generated specifications for APR. Experimental results show that SpecTune improves fault localization and APR effectiveness than the baselines.
arXiv:2604.11943v3 Announce Type: replace Abstract: An OS kernel that runs LLM inference internally can read the model's own next-token logit distribution before any text is generated, and act on it as a governance primitive. I present ProbeLogits, a kernel-level operation that performs a single forward pass and reads specific token logits to classify an agent's action as safe or dangerous, with zero learned parameters. Because the probe reads a logit from the same base model the agent already runs, it removes the second model a fine-tuned guard requires: the marginal cost of a safety check becomes a single logit read. I evaluate ProbeLogits on three base models (Qwen2.5-7B, Llama-3-8B, Mistral-7B) across three external benchmarks (HarmBench, XSTest, ToxicChat). On HarmBench non-copyright, all three reach a 97-99% block rate. On ToxicChat (n=1,000), ProbeLogits attains F1 parity-or-better against Llama Guard 3: Qwen2.5-7B Safe/Dangerous reaches F1=0.812 (+13.7 pp, bootstrap 95% CIs disjoint), Llama-3 matches within CI (+0.4 pp), and Mistral exceeds by +4.4 pp. Classification is a measured 2.4-3.4x faster than Llama Guard 3 (332-556 ms vs. 851-1,142 ms), because it reads a single logit position instead of generating tokens. A calibration strength alpha acts as a deployment-time policy knob rather than a learned hyperparameter, trading recall for precision per operation class. I implement ProbeLogits within Anima OS, a bare-metal x86-64 kernel written in ~285,000 lines of Rust. Because agent actions must pass through kernel-mediated host functions, enforcement operates below the WASM sandbox boundary, making it substantially harder to circumvent than application-layer classifiers.
arXiv:2604.13416v3 Announce Type: replace Abstract: Advances in radiance fields have enabled photorealistic novel view synthesis. In several domains, large-scale real-world datasets have been developed to support comprehensive benchmarking and to facilitate progress beyond scene-specific reconstruction. However, for distractor-free radiance fields, a large-scale dataset with clean and cluttered images per scene remains lacking, limiting the development. To address this gap, we introduce DF3DV-1K, a large-scale real-world dataset comprising 1,048 scenes, each providing clean and cluttered image sets for benchmarking. In total, the dataset contains 89,924 images captured using consumer cameras to mimic casual capture, spanning 128 distractor types and 161 scene themes across indoor and outdoor environments. A curated subset of 41 scenes, DF3DV-41, is systematically designed to evaluate the robustness of distractor-free radiance field methods under challenging scenarios. Using DF3DV-1K, we benchmark nine recent distractor-free radiance field methods and 3D Gaussian Splatting, identifying the most robust methods and the most challenging scenarios. Beyond benchmarking, we demonstrate an application of DF3DV-1K by fine-tuning a diffusion-based 2D enhancer to improve radiance field methods, achieving average improvements of 0.96 dB PSNR and 0.057 LPIPS on the held-out set (e.g., DF3DV-41) and the On-the-go dataset. We hope DF3DV-1K facilitates the development of distractor-free vision and promotes progress beyond scene-specific approaches. The dataset and leaderboard are available at https://johnnylu305.github.io/df3dv1k_web/.
arXiv:2501.10711v5 Announce Type: replace Abstract: Code-related benchmarks play a critical role in evaluating large language models (LLMs), yet their quality fundamentally shapes how the community interprets model capabilities. In the past few years, awareness of benchmark quality has grown. Yet, after a decade-scale (2014-2025) survey over 672 code benchmarks, we observed a lag between growing awareness and actual practice. For example, in 2025 alone, the number of benchmarks that ignore code coverage when providing test cases nearly matches the total count accumulated across the previous ten years. In response, we take a clear position: Code benchmarks must prioritize rigor in benchmark construction, reliability in evaluation, and reproducibility in release. To operationalize this position, we introduce a code benchmark guideline HOW2BENCH with 55 checklists. Finally, our further human study also exposed that the current issues not only stem from the significant effort required, but also from a lack of awareness regarding their importance.
arXiv:2607.04108v1 Announce Type: new Abstract: Large language models are increasingly used as evolutionary engines for scientific discovery: generate candidates, select winners, feed them back as parents, and repeat. We audit whether this loop actually compounds discovery in scientific equation discovery, a setting where finite samples make structure underdetermined and interpolation easy. Under matched LLM-call budgets, parent-conditioned evolution is indistinguishable from fresh independent sampling: median OOD NMSE is 0.045 vs. 0.049, instructed multi-parent crossover is worse, final success is predicted by initial proposal quality, and multiple iteration schemes fail to add solved problems. Operationally, the loop reduces to what it produces: a dictionary of candidate terms. We turn that diagnosis into PTB-Search, a one-generation method for componentized scientific discovery. PTB-Search samples independent LLM proposals once, extracts reusable terms into a per-problem dictionary, and performs train-only set-level sparse selection with least-squares coefficients. Its central principle is that underdetermined data identifies the joint behavior of term sets, not reliable per-term credit. On identical dictionaries and zero additional LLM calls, set-level selectors solve 165--169 of 717 cells, while single-term reductions solve only 74--78. On the official 239-problem LLM-SRBench split, PTB-Search reaches 73.2% Acc0.1 with Llama-3.1-8B and 77.0% with a single-seed DeepSeek-V4 anchor, versus 49.2% for the best reported baseline, using one tenth of the standardized call budget. A program-domain stress test gives a scoped boundary: generation count remains unreliable, while retained external state can help in harder non-linear spaces. Across these results, LLMs are best understood as material suppliers; discovery is carried by external set-level selection over reusable components.
arXiv:2604.14037v2 Announce Type: replace Abstract: Parameter space is not function space for neural network architectures. This fact, investigated as early as the 1990s under terms such as ``reverse engineering," or ``parameter identifiability", has led to the natural question of parameter space symmetries\textemdash the study of distinct parameters in neural architectures which realize the same function. Indeed, the quotient space obtained by identifying parameters giving rise to the same function, called the \textit{neuromanifold}, has been shown in some cases to have rich geometric properties, impacting optimization dynamics. Thus far, techniques towards complete classifications have required the analyticity of the activation function, notably excising the important case of ReLU. Here, in contrast, we exploit the non-differentiability of the ReLU activation to provide a complete classification of the symmetries in the shallow case.
arXiv:2604.14221v2 Announce Type: replace Abstract: Reliable evaluation of anomaly detection methods in multivariate time series remains an open challenge, largely due to the limitations of existing benchmark datasets. Current resources often lack fine-grained anomaly annotations, do not provide explicit intervariable and temporal dependencies, and offer little insight into the underlying generative mechanisms. These shortcomings hinder the development and rigorous comparison of detection models, especially those targeting interpretable and variable-specific outputs. To address this gap, we introduce Fun-TSG, a fully customizable time series generator designed to support high-quality evaluation of anomaly detection systems. Our tool enables both fully automated generation, based on randomly sampled dependency structures and anomaly types, and manual generation through user-defined equations and anomaly configurations. In both cases, it provides full transparency over the data generation process, including access to ground-truth anomaly labels at the variable and timestamp levels. Fun-TSG supports the creation of diverse, interpretable, and reproducible benchmarking scenarios, enabling fine-grained performance analysis for both classical and modern anomaly detection models.