arXiv:2607.12101v1 Announce Type: new
Abstract: OpenAPI specifications are the primary input for black-box testing tools in microservice systems (MSS), yet prior work shows these specifications are often incomplete, inconsistent, or incorrect. Despite this, most studies on OpenAPI-based black-box testing assume correct specifications and evaluate tool performance. We address this gap by introducing a literature-grounded taxonomy of six OpenAPI specification fault classes. We inject faults at five severity levels, and evaluate the resulting mutated specifications on two microservice benchmarks, TrainTicket and SocialNetwork, using three testing tools: EvoMaster, RESTler, and Schemathesis. We measure the impact of these faults using code coverage, specification coverage, request/response quality, and behavioral diversity. Our results show that specification faults cause strong and heterogeneous degradation patterns across testing tools and systems. Faults in method semantics cause broad degradation across all metrics, while others, such as modifications to response codes, remain weak. Relaxations of schema constraints cause hidden degradation, with no impact on code and specification coverage but a large impact on request/response quality. These findings demonstrate that specification quality directly shapes black-box API testing effectiveness. Also, code and specification coverage-only evaluations can understate the impact of specification faults on black-box testing in MSS and should be complemented by request/response quality and behavioral diversity.
Science Journals
arXiv:2605.10834v2 Announce Type: replace
Abstract: AI pentesting agents are increasingly credible as offensive security systems, but current benchmarks still provide limited guidance on which will perform best in real-world targets. Existing evaluation protocols assess and optimize for predefined goals such as capture-the-flag, remote code execution, exploit reproduction, or trajectory similarity, in simplified or narrow settings. These tools are valuable for measuring bounded capabilities, yet they do not adequately capture the complexity, open-ended exploration, and strategic decision-making required in realistic pentesting. In this paper, we present a practical evaluation protocol that shifts assessment from task completion to validated vulnerability discovery, allowing evaluation in sufficiently complex targets spanning multiple attack surfaces and vulnerability classes. The protocol combines structured ground-truth with LLM-based semantic matching to identify vulnerabilities, bipartite resolution to score findings under realistic ambiguity, continuous ground-truth maintenance, repeated and cumulative evaluation of stochastic agents, efficiency metrics, and reduced-suite selection for sustainable experimentation. This protocol extends the state of the art by enabling a more realistic, operationally informative comparison of AI pentesting agents. To enable reproducibility, we also release expert-annotated ground truth and code for the proposed evaluation protocol: https://github.com/ethiack/ethibench.
arXiv:2605.12765v3 Announce Type: replace
Abstract: Large Language Models memorize vast amounts of training data, raising concerns regarding privacy, copyright infringement, and safety. Machine unlearning seeks to remove the influence of a targeted forget set while preserving model performance, ideally approximating a model retrained from scratch without the forget set. Existing approaches aim to achieve this by updating model parameters via gradient-based methods. However, these updates are computationally expensive, lead to irreversible weight changes, and degrade when the model is quantized for deployment. A recent alternative to changing model weights is activation engineering, where activations are changed during inference to steer model behavior. Despite circumventing weight editing, naive activation steering introduces its own failure modes, as a single global steering vector applies the same intervention to every input, leading to unintended changes in model behavior. We introduce Inference-Time Unlearning via Gated Activation Redirection (GUARD-IT), a training- and gradient-free method that unlearns via input-dependent activation steering at inference time. The resulting intervention is applied as a norm-preserving rotation in the residual stream, leaving model weights untouched. Experiments on TOFU and MUSE show that GUARD-IT matches or exceeds 12 gradient-based baselines across three model scales, while being the only method to simultaneously preserve utility, suppress memorization, and avoid catastrophic collapse across all settings. GUARD-IT further supports continual unlearning without retraining, and remains effective under quantization, a scenario in which parameter-editing methods degrade.
arXiv:2605.15222v2 Announce Type: replace
Abstract: Large language models (LLMs) can often generate functionally correct code, but their ability to produce efficient implementations for performance-critical systems tasks remains limited. Existing code benchmarks mainly emphasize correctness or algorithmic problem solving, while realistic systems-level optimization is still underexplored. To address this gap, we introduce PerfCodeBench, an executable benchmark for evaluating LLMs on high-performance code optimization. The tasks require system-level implementation choices, hardware-aware optimization, and careful handling of performance bottlenecks. Each task includes executable correctness checks, a baseline implementation, and a reference optimized solution. This allows us to evaluate both correctness and runtime-oriented efficiency. Our evaluation on a broad set of state-of-the-art LLMs shows a clear gap between model-generated code and expert-optimized implementations. The gap is especially large on tasks involving parallelism and GPU operations. Current models also show weaknesses in cross-language robustness and in consistently reaching expert-level efficiency. These results suggest that performance-aware evaluation are still needed. LLMs should move beyond generating merely correct code toward producing efficient systems software. We submit the benchmark data, evaluation infrastructure, and complete logs of all LLMs-generated code at https://anonymous.4open.science/r/perfcodebench-7CDE.
arXiv:2605.28198v2 Announce Type: replace
Abstract: Existing approaches for synthetic tabular data generation are based on either purely generative models or LLMs, both of which struggle with data heterogeneity, logical consistency, rare-event coverage, and robustness in low-data regimes. In this paper, we propose a hierarchical hybrid top-down and bottom-up (H-TDBU) framework that decouples semantic structures from stochastic texture. In the top-down path, structure-driven logical constraints and cross-modal alignment rules are constructed, while in the bottom-up path, lightweight tabular generators are used to learn local statistical patterns from real data. The two paths are consolidated in a unified synthesis engine with an iterative feedback loop. We evaluate the framework on weak multimodal financial benchmarks combining tabular and sentiment-text data. Experimental results show that our H-TDBU approach improves train-synthetic-test-real performance over neural baseline methods while preserving semantic consistency. Our results suggest that hierarchical rule-guided synthesis provides an effective mechanism for combining controllability, semantic coherence, and statistical fidelity in synthetic data generation.
arXiv:2606.00190v2 Announce Type: replace
Abstract: Social deduction games, or hidden-role games, are multiplayer games in which players are assigned private roles and act under asymmetric information about other players' roles and actions. In the canonical example Werewolf, werewolves conceal their roles and mislead the other players, while the seer can obtain role information about a chosen player. Thus, a central functionality of such games is controlling which players can access which information. In typical play, this control is implemented by a trusted human moderator, who assigns roles, mediates secret actions, and reveals outcomes. This reliance raises the barrier to participation and introduces a trusted third party as a single point of failure. In this work, we show that Werewolf can be played without a moderator or any digital device, using only ordinary playing cards. Our construction maintains a shared pool of cards that is observable to all players and manipulated according to a common public procedure, while its interpretation depends on each player's private role. This induces role-dependent views from a single public sequence of card operations. Consequently, even without private messages, werewolves can identify one another and coordinate, and the seer can test whether a chosen player is a werewolf in each round. The proposed implementation is built from card-based physical cryptographic primitives, such as face-down commitments and verifiable shuffles, and higher-level subprotocols for intra-role information sharing, secret action designation, and attribute testing. These subprotocols implement the moderator's core functions while keeping all card operations public and auditable under standard assumptions on physical card operations. We show that the resulting complete moderatorless implementation of Werewolf scales to an arbitrary number n of players using O(n^3) cards.
arXiv:2606.02029v2 Announce Type: replace
Abstract: We study online scheduling with obligatory testing on $m$ identical parallel machines, with the objective of minimizing the sum of completion times. Each job comprises a test of known length and a processing operation of initially unknown length. The processing time is revealed only when the test completes. Unlike in optional testing models, the scheduler does not choose whether to acquire information. Instead, it must decide how to allocate machine capacity between testing unrevealed jobs and processing jobs whose sizes are already known. Previous single-machine lower-bound constructions suggest a natural $\sqrt{2}$ benchmark [ESA 2024: 48:1-14]. However, these constructions cannot be directly transferred to identical parallel machines by a simple replication argument. An online algorithm may interleave jobs from different copies, and the test and processing operation of a job need not be scheduled on the same machine. We address this difficulty by introducing a completion-threshold framework that reasons directly about global progress under total machine capacity. For each $X$, let $T_X$ be the earliest time at which the algorithm has completed at least $X$ jobs. The identity $\sum_{X=1}^{N}T_X$ then converts pointwise progress bounds into lower bounds on the total completion time. Using this framework, we prove a three-type lower bound of $1.4811$ and a dyadic multi-type lower bound tending to $3/2$. The latter also improves the deterministic single-machine lower bound from $\sqrt{2}$ to $3/2$. On the algorithmic side, we give a parallel version of single-machine 1-SORT and prove that, if single-machine 1-SORT is $\rho$-competitive, then its parallel version is $\frac{2(m+\rho-1)}{m+1}$-competitive on $m$ identical machines.
arXiv:2606.03363v3 Announce Type: replace
Abstract: Text-to-SQL enables natural language access to databases, and recent LLMs have substantially advanced its capabilities. Existing benchmarks such as Spider, BIRD, and Spider~2.0 evaluate schema generalization, large-scale databases, and realistic workflows, but largely overlook enterprise scenarios where SQL generation depends on private business knowledge, such as internal metrics, reporting conventions, and organizational rules. We introduce EntSQL, an enterprise-oriented Text-to-SQL benchmark for evaluating long-context grounding over proprietary business documents. EntSQL contains 1,066 aligned Chinese-English semantic examples across five business domains, with most examples requiring domain knowledge beyond the question and schema and involving complex SQL structures. On English inputs, the best evaluated system reaches only 15.9\% when long-form documents are provided, highlighting the difficulty of grounding SQL generation in enterprise knowledge.
arXiv:2606.03691v2 Announce Type: replace
Abstract: The Formalized Process Description (FPD) according to VDI/VDE 3682 provides a standardized graphical notation for describing processes across engineering domains but lacks a standardized, tool-independent data format for machine-readable model exchange. This paper presents an AutomationML (AML) domain library that formalizes the complete set of FPD language elements, their attributes, connection semantics, and graphical representation information as class libraries based on the Computer Aided Engineering Exchange (CAEX) 3.0 metamodel. The library comprises five interrelated parts: a RoleClassLib defining the semantic roles, an InterfaceClassLib for connection types, two AttributeTypeLibs for the information model and diagram interchange, and a SystemUnitClassLib providing instantiation templates. Key design decisions regarding inheritance, diagram structure, hierarchical decomposition, and the representation of graphical information are discussed along with the alternatives that were considered. A bidirectional mapping tool demonstrates the library's applicability by converting between a web-based FPD modeler and AML.
arXiv:2606.03933v2 Announce Type: replace
Abstract: Structural Health Monitoring (SHM) aims at the real-time monitoring of the integrity of engineering structures, with Guided-waves (GWs) providing high sensitivity to damage presence and to ageing effects for thin-walled components. In conventional GW-based SHM, a bonded piezoelectric transducer (PZT) emits a short tone burst that produces an Initial Wave Packet (IWP) propagating through the structure. As this packet interacts with boundaries and potential damages, additional scattered wave packets are produced. A major limitation of such approaches lies in the simultaneous excitation of multiple dispersive GW modes by a single PZT, which significantly complicates signal interpretation and damage monitoring. In this context, this work proposes the Physics-Informed Single Atom Convolutional Matching Pursuit (PISACMP) method, a signal decomposition method grounded in the physical principles governing wave propagation. In contrast with purely data-driven or numerically intensive techniques, the proposed approach embeds strong physical constraints into a low-dimensional and computationally efficient signal representation. This formulation enables the direct identification of key physically meaningful features, including modal wavenumber functions and propagation distances between actuator, damage and sensors. These extracted features, especially source-damage-sensor distances, allows to subsequently perform damage location using well established Elliptical Localization techniques. The principal novelty of this study lies in integrating wave propagation physics into a compact signal decomposition framework and developing an interpretable damage localization methodology for GW-SHM applications.
arXiv:2606.04669v2 Announce Type: replace
Abstract: Achieving quantum resilience in software systems is increasingly critical as quantum computing threatens conventional cryptographic mechanisms. One approach to achieving quantum resilience is implementing software systems with Post-Quantum Cryptography (PQC) mechanisms. However, developers and organisations face significant challenges in integrating PQC into real world systems due to implementation complexity, limited expertise, and insufficient organisational readiness. Existing studies largely emphasise cryptographic performance and algorithmic security, offering limited insight into the broader Human Organisation Technology (HOT) factors influencing successful PQC implementation. This Systematisation of Knowledge (SoK) examines PQC implementation approaches and challenges through the Human, Organisation, and Technology (HOT) dimensions. Our analysis reveals a strong technological bias in current research, with human and organisational factors remaining underexplored. We further show that PQC implementation challenges are interconnected constraints spanning all HOT dimensions. These findings suggest that PQC implementation extends beyond cryptographic replacement and requires coordinated HOT transformation to achieve quantum resilience. To support this, we propose the PQC HOT framework, a conceptual framework that synthesises implementation approaches and challenges to support structured analysis. We outline future research directions and design implications for scalable and sustainable quantum-resilient software systems.
arXiv:2606.08167v2 Announce Type: replace
Abstract: Recent research has established empirical scaling laws to predict model performance on multi-domain data mixtures. However, a theoretical understanding of these model loss behaviors remains absent. In this work, we propose a unified framework to explain the underlying mechanics of data mixing. Our approach extends theoretical perspectives originally developed for standard neural scaling laws (e.g., Kaplan and Chinchilla) to the multi-domain setting. Based on the distributional assumption that domains overlap on fundamental skills while diverging on specialized skills, we identify two key factors that govern the domain losses of models trained on different data mixtures: \textit{Capacity Competition}, where the allocation of finite model capacity couples domain losses globally, and \textit{Noise Reduction}, where optimal weights shift toward harder-to-learn domains to minimize overall noise. Empirical evaluations show that our framework outperforms existing baselines by fitting the loss landscape with a lower Mean Relative Error and identifying higher-performing training mixtures. Most importantly, our model successfully extrapolates across scales, predicting highly effective mixtures for large, unseen scales using parameters fitted on smaller ones. In addition, our model achieves these results using significantly fewer parameters compared to previous empirical laws. Our code is available at https://github.com/meiqwq/Explaining-Data-Mixing-Scaling-Laws.
arXiv:2606.09076v3 Announce Type: replace
Abstract: Reward models are central to text-to-image post-training, but visual preference is subjective and better represented as a distribution over rubric scores than as a deterministic scalar. Existing scalar, score-token, and pairwise reward models over-compress uncertainty and fine-grained score differences, while reasoning-based generative rewards provide stronger judgments but are costly to deploy and difficult to use as direct optimization signals. We propose Z-Reward, a teacher-student reward modeling framework that decouples reasoning-heavy judgment from efficient reward deployment. The teacher is a large VLM that uses reasoning to infer rubric-aligned score distributions, and is trained with Group-wise Direct Score Optimization (GDSO), which combines policy-gradient rewards from distribution expectations with direct pointwise and pairwise supervision on score distributions and score gaps. The student is trained with Reasoning-Internalized Score Distillation (RISD), which transfers the teacher's reasoning-conditioned score distribution into a compact VLM without requiring explicit reasoning chains at inference time. On our internally annotated evaluation set, the 27B GDSO teacher reaches 89.6% human preference accuracy, outperforming SFT, RewardDance, and GRPO, while the 9B RISD student reaches 88.6%, outperforming the OPD baseline and closely matching the larger teacher. We further show that Z-Reward can serve as a differentiable reward signal for text-to-image optimization, yielding a 41.3% net human-preference improvement over the SFT baseline.
arXiv:2606.11520v4 Announce Type: replace
Abstract: Training capable OS agents requires data that simultaneously captures structured user intents, multi-turn task delegation, and grounded tool execution--properties absent from existing datasets. We propose ISE (Intent -> Simulate -> Execute), a three-stage synthesis paradigm that addresses these gaps jointly. Stage 1 constructs roughly 50000 structured intents via a 4D framework (Persona x Domain x Task x Complexity); after deduplication the pool contains 43956 unique intents and attains a Vendi Score of 61.57 over the entire pool on mpnet-base-v2 embeddings (cosine kernel, q=1). Stage 2 drives multi-turn user-agent interaction through a role-locked user simulator that grounds each user turn in actual execution outcomes, producing 23132 complete trajectories averaging 8.12 user turns and 68.24 total dialogue turns. Stage 3 runs every tool call inside a live, isolated OS workspace, generating authentic failure-recovery dynamics instead of simulated responses. Fine-tuning on ISETrace improves ClawEval pass@1 from 19.3 to 37.7 using Qwen3-8B on agent tool-use tasks with a standard protocol. This result outperforms zero-shot GPT-4o and the larger Qwen3-32B base model which is four times bigger. An ablation on Stage 2 proves multi-turn simulation brings a large portion of the performance gain. We release all source code and dataset at https://github.com/Valiere01/ISE-Trace.
arXiv:2606.12634v3 Announce Type: replace
Abstract: Long-horizon tool-use reinforcement learning learns from outcome verification, but trajectory-level advantages are broadcast over reasoning, API, and answer tokens. Direct self-distillation can supply a denser signal, but in our experiments it can also destroy tool use by rehearsing teacher behavior without identifying which actions the verifier rewards. We introduce Sibling-Guided Credit Distillation (SGCD), which uses distillation for bounded credit weighting rather than as a competing actor loss. Dynamic sampling produces mixed successful and failed sibling rollouts; an external LLM summarizes their contrast into a training-only credit reference; and detached teacher/student divergence reshapes GRPO token advantages. The deployed student receives only the clean task prompt. Across AppWorld and tau^3-airline, SGCD reports higher held-out point estimates than GRPO-family comparators: AppWorld TGC improves from 42.9 to 45.6 on test_normal and from 24.7 to 27.0 on test_challenge, and tau^3-airline held-out evaluator score improves from 0.583 to 0.602. These results support a narrow design rule for long-horizon tool-use agents: use distillation to guide credit assignment while keeping policy gradient in charge of the actor update.
arXiv:2607.11905v1 Announce Type: new
Abstract: The Indian Summer Monsoon (ISM) represents one of the most consequential and dynamically complex phenomena in the global climate system, yet its prediction remains challenging for both physics-based and data-driven models. This study evaluates Google's GraphCast, a machine-learning-based global weather prediction system, against ERA5 reanalysis and IMERG precipitation observations for the boreal summer season (June-September, JJAS) over 2021-2024. We analyze deterministic GraphCast forecasts initialized at 00 UTC using four 6-hourly lead times, composited for +24 h, +48 h, and +72 h lead performance across the ISM domain. The evaluation includes climatological mean state, rainfall intensity distribution, thermodynamic drivers (DTT, Q1, Q2), monsoon dynamics, and variability across multiple timescales. Results show GraphCast reproduces the broad spatial pattern of monsoon rainfall and the annual cycle with good fidelity at short leads, but exhibits a domain-averaged wet bias over the core monsoon region and strong suppression of rainfall variance across nearly all timescales (regional power-spectrum variance ratio ~0.14 relative to IMERG). The rainfall intensity distribution is right-shifted and compressed, with moderate-heavy rainfall (95th percentile) biased wet and extreme events underrepresented. These biases are accompanied by a deficient lower-tropospheric Q1 profile and degraded northward-propagating intraseasonal variability. Overall, GraphCast shows a deterministic smoothing signature in precipitation, providing key diagnostics for next-generation AI weather models and their use in tropical extended-range forecasting.
arXiv:2607.12293v1 Announce Type: new
Abstract: Predicting pedestrian crossing intention is a safety-critical task for autonomous driving, yet existing approaches often rely on single-modal inputs or dense multimodal fusion strategies that inadequately capture complementary visual and kinematic information while introducing redundant inter-modal interactions. We propose ADAPT (Adaptive Domain-Aware Pedestrian Crossing Transformer), a multimodal framework that jointly models local and global visual context together with temporal motion dynamics for accurate pedestrian crossing intention prediction. ADAPT processes four spatially aligned visual modalities, including RGB images, local depth maps, global semantic maps, and global depth maps, together with ego-vehicle speed, pedestrian bounding boxes, and skeleton pose information through five specialized modules: a weight-shared Swin Transformer V2 backbone for visual feature extraction, a Cross-Modality Guided Attention module for hierarchical visual fusion, a Mamba-based Motion Feature Encoding module for efficient temporal modeling, a Sparse Cross-Modal Attention module that selectively preserves the most informative inter-modal interactions, and a Vision Transformer-based Temporal Feature Fusion module for sequence-level prediction. Extensive experiments on the JAAD and PIE benchmark datasets demonstrate that ADAPT consistently outperforms existing state-of-the-art methods while maintaining low computational complexity. On JAAD, the proposed method achieves an AUC of 0.73 on JAADbeh and 0.85 on JAADall, while on PIE it achieves an accuracy of 0.92 and an AUC of 0.90. Furthermore, ADAPT performs inference in only 17.23 ms per sample, offering an effective balance between predictive accuracy and real-time deployment efficiency for intelligent transportation and autonomous driving applications.
arXiv:2607.12328v1 Announce Type: new
Abstract: Next-generation low Earth orbit (LEO) satellite constellations face critical bottlenecks in spectral efficiency and onboard hardware complexity. To overcome these limitations, this paper introduces a novel architecture enabled by metasurface antennas (MAs) at the LEO satellites. In particular, MAs are metasurface-integrated feed antennas that perform high-precision beamforming directly in the wave domain, thereby effectively mitigating multi-user interference. Based on such an antenna architecture, a weighted sum rate (WSR) maximization problem is formulated by jointly optimizing the scheduling of feed antennas to terrestrial users (TUs) and the passive beamforming of the metasurface for system performance enhancement. To address this mixed-integer nonlinear programming (MINLP) challenge, an alternating optimization (AO)-based joint scheduling and beamforming algorithm is proposed. On the one hand, the proposed algorithm incorporates a polynomial-time minimum-cost maximum-flow (MCMF) method, which is dedicated to the optimal scheduling of feed antennas and TUs. On the other hand, it adopts a weighted minimum mean square error (WMMSE) method integrated with semidefinite relaxation (SDR) technique, which is tailored for metasurface beamforming design. Simulation results confirm the effectiveness of the proposed algorithm for MA-enabled LEO satellite constellation communications.
arXiv:2607.12332v1 Announce Type: new
Abstract: We study the gradient flow dynamics of diagonal linear networks for regression tasks under infinitesimal initialization. Extending Theorem 1 from Pesme & Flammarion (2023), we generalize the analysis to both deep diagonal linear networks and a broader class of two-layer diagonal linear networks (as defined in Definition 4.1). Specifically, we demonstrate that the training trajectories of these models can be equivalently characterized by the proposed Algorithm 1. We further prove that this algorithm converges to the solution of a modified $ \mathcal{l}_1 $ norm minimization problem. As a result, we establish that the implicit bias of both network architectures corresponds to a modified $ \mathcal{l}_1 $ norm in the regime of infinitesimal initialization. Additionally, we provide insights into the underlying mechanisms governing these dynamics by identifying the Structural Invariant Manifold (SIM) (Zhao et al., 2026) as the key geometric structure that shapes the learning process.
arXiv:2605.27947v2 Announce Type: replace
Abstract: World Action Models (WAMs) improve robot manipulation by using video-based future representations to condition action generation. In pixel-space WAMs, however, the best action condition is not necessarily the fully denoised video. Controlled denoising-depth scans show that video refinement can reduce action error up to a state-dependent point, after which the gain may saturate or even reverse when late predictions become less action-relevant or physically unreliable. This suggests that action generation should use a state-dependent point along the video noise trajectory rather than a fixed terminal denoising depth. We introduce State-Adaptive Noise Trajectory Scheduler (SANTS), a lightweight scheduler for video-to-action diffusion policies. At each video decision point, SANTS reads the current video-state representation and noise level, then jointly predicts a cumulative stopping hazard and a relative noise-progression ratio. SANTS is post-trained with a path-level reward computed after the frozen action branch generates the final action chunk, so the scheduler is optimized for downstream action quality rather than intermediate video fidelity, while redundant video-state updates are explicitly penalized. Experiments show that SANTS reaches \(94.4\%\) overall success on RoboTwin 2.0 and \(73.1\%\) average success across seven real-robot tasks, while reducing latency by \(81.7\%\) and \(79.0\%\) relative to full video denoising, respectively. These results indicate that adaptive selection along the video noise trajectory can preserve the control benefits of WAM-style future reasoning while removing much of its redundant inference cost.
arXiv:2606.13460v2 Announce Type: replace
Abstract: Semantic 3D occupancy provides a voxelized world state for autonomous driving and robot decision making, but object and rare-class errors can affect free-space interpretation, collision checking, and temporal state propagation. We show that a common VLM strategy, aligning 3D voxel or object features with crop-caption embeddings, improves text-space similarity without reliably improving closed-set occupancy mIoU. Motivated by this mismatch, we propose VISA, a training-time semantic auditing approach for existing occupancy world models. VISA queries an offline VLM on a representative crop of each physical object instance, obtains a structured audit with class hypotheses, plausible confusions, reliability, attributes, and evidence, and propagates it along the object track. The audit is grounded to matched 3D object voxels and distilled into semantic logits through reliability-weighted taxonomy, attribute-factor, and scene-level audit graph losses, while inference remains unchanged and requires no VLM. On nuScenes, averaged across three runs, VISA improves OccWorld from 19.06 to 20.05 mIoU and GaussianWorld from 21.36 to 21.91 mIoU; on GaussianWorld, object mIoU improves from 18.18 to 19.16 and rare-class mIoU from 15.60 to 16.79. These results suggest that VLMs are better suited to closed-set occupancy as reliability-aware semantic auditors than as generic caption-embedding targets.
arXiv:2605.31437v3 Announce Type: replace
Abstract: Interpreting computed tomography (CT) requires review of hundreds of volumetric slices and remains time-intensive and expertise-dependent. Automated CT report generation offers a promising route to improving clinical efficiency, yet the field still lacks a generalizable CT report generation foundation model that supports multi-region reporting and remains robust across external real-world cohorts. Intrinsic inconsistencies in reporting style and diagnostic terminology across cohorts make naive joint training difficult. Here we present Astra, a generalizable CT report generation foundation model developed on 90,678 thoracoabdominal CT-report pairs collected from five sites worldwide (CTRgDB), comprising 353,671 abnormalities spanning eight organ systems. By harmonizing report style and further refining diagnostic consistency via reinforcement learning, Astra achieves style-consistent and diagnostically accurate report generation across diverse anatomical regions and institutions. Evaluated on CTRgDB and six external cohorts, Astra achieves state-of-the-art performance with a 38.4% average improvement in fine-grained diagnostic metrics (P<0.001). Deployed at external clinical sites without any site-specific fine-tuning, Astra accelerated chest report drafting by 29.6% and improved abdominal report completeness by 11.3% among junior and mid-level radiologists (P<0.001). Furthermore, Astra demonstrates broad utility as a foundation for CT AI development, improving downstream diagnostic performance and scaling vision-language pretrain through high-quality report synthesis. Overall, Astra serves as a broadly accessible clinical assistant and a pivotal infrastructure for the next generation of AI-powered healthcare. The code for Astra is publicly available at https://github.com/zh-Wang-Med/Astra.
arXiv:2606.17253v5 Announce Type: replace
Abstract: Large Language Models and vision-language models have shown remarkable success in the front-end design of Very Large-Scale Integrated Circuits, yet their capabilities for VLSI physical design remain significantly underexplored. The primary cause is the lack of standardized benchmarks for evaluating agentic physical design workflows that require high-dimensional, multi-stage optimization under strict design constraints, coordinated interaction with diverse Electronic Design Automation tools, and iterative refinement. This work introduces PDAGENT-BENCH, a comprehensive and multi-dimensional benchmark for evaluating LLM/VLM-based agents across the physical design stack. PDAGENT-BENCH integrates both task-level assessment and workflow-level execution. The benchmark suite contains 353 curated problems that combine conceptual questions with real-world industrial artifacts, with expert-validated references and executable solutions. In addition, the benchmark provides a unified, human-aligned agentic physical design workflow framework that enables closed-loop evaluation of holistic physical design in realistic EDA environments. Experiments on 11 state-of-the-art models reveal that while modern LLMs/VLMs perform competitively on conceptual tasks, they remain substantially limited in tool-centric execution (e.g., 42.2% on Innovus script generation) and long-horizon, multi-stage reasoning. Our studies further show that human-skill-enhanced agentic workflows significantly improve end-to-end physical design performance. PDAGENT-BENCH establishes a standardized, reproducible, and realistic evaluation framework for advancing LLM/VLM-driven holistic physical design automation. To ensure full reproducibility and broad accessibility, we will release PDAgent-Bench together with its agentic workflow framework, instantiated on open-source PDKs (e.g., Nangate45, ASAP7) and open EDA tools (e.g., OpenROAD).
arXiv:2606.20101v3 Announce Type: replace
Abstract: Audio editing aims to modify specific content in an existing audio clip according to a text instruction or description while preserving the remaining acoustic content. Despite the remarkable progress of diffusion models, existing training-based editing methods mainly rely on the local inductive biases and cross-attention interaction in convolutional U-Net backbones, which often hinder long-range semantic alignment and precise understanding and localization of instructions. In contrast, diffusion transformers provide stronger global modeling and multimodal fusion, but existing editing architectures usually adopt a simple stack of diffusion transformer blocks. Applying joint attention over concatenated audio and text tokens in all blocks results in quadratic complexity with respect to token length. To balance editing performance and efficiency, we propose a novel instruction-guided audio editing framework based on rectified flow matching (RFM), named RFM-Editing 2, built on a hybrid two-stage diffusion transformer. The proposed model performs joint attention over audio and text tokens to establish coarse semantic alignment at the low-resolution stage, then switches to alternating joint-attention and cross-attention blocks to refine editing details at the high-resolution stage. This coarse-to-fine strategy enables efficient and accurate instruction-guided audio editing. Experiments show that the proposed framework achieves notable performance gains on challenging editing tasks involving overlapping audio events and complex instructions, while substantially improving editing efficiency.
arXiv:2607.11896v1 Announce Type: new
Abstract: Forecasting particulate matter (PM10) requires both station-scale accuracy and continuous spatial fields, especially during severe dust storms. Chemical transport models (CTMs) provide gridded forecasts but retain local biases, whereas graph neural networks (GNNs) track monitoring sites well at short lead times but do not produce gridded outputs. Here we present OmniPM-Net, a Convolutional Conditional Neural Process (ConvCNP)-based fusion model that reconciles these two forecast types within a shared spatial representation. A terrain-aware Gaussian set convolution lifts irregular GNN station forecasts onto a regular grid, where a multi-scale Spatial Source Attention (SSA) module blends them with Copernicus Atmosphere Monitoring Service (CAMS) forecasts; a shared omni-query readout then decodes this representation into consistent PM10 predictions at either stations or grid cells over a 108 h horizon. Evaluated across 1,618 air-quality monitoring stations throughout China over the full year of 2024, OmniPM-Net matches the station-level accuracy of the stronger GNN baseline (mean absolute error 21.14 versus 22.00 ug/m3) and reduces the CAMS mean absolute error by 30%, while simultaneously delivering the gridded fields that the discrete GNN cannot. Its clearest gains are in the high-concentration tail, where the 90th-percentile MAE falls by 9% relative to the GNN and 25% relative to CAMS, and during dust episodes, where it improves categorical detection skill while tracking the evolving spatial trajectory.