arXiv:2602.11454v3 Announce Type: replace
Abstract: We study $\left(\epsilon,\delta\right)$-differentially private algorithms for the problem of approximately computing the top singular vector of a matrix $A\in\mathbb{R}^{n\times d}$ where each row of $A$ is a data point in $\mathbb{R}^{d}$. Following Dwork-Talwar-Thakurta-Zhang (STOC 2014), we consider the privacy model where neighboring inputs differ by one single row. We give a novel algorithm that achieves beyond-worst-case guarantees for input matrices with low coherence, which is a structural property of matrices in many applications, including but not limited to i.i.d. data. Our algorithm contributes to the extensive literature on private power iteration methods, where we introduce a new filtering technique which adapts to this coherence parameter. Our work departs from and complements the work by Hardt-Roth (STOC 2013) which achieves beyond-worst-case guarantees for the more restrictive privacy model where neighboring inputs differ in one single entry by at most 1.
Science Journals
arXiv:2605.19774v1 Announce Type: new
Abstract: The breakup of thinning (stretching) liquid ligaments is strongly influenced by localized perturbations arising from impurities or suspended particles. Using numerical simulations and analytical modelling, we investigate the role of a solid particle on the breakup dynamics of a stretching liquid ligament. We show that particle-induced perturbations trigger a universal pinch-off dynamics in the viscous regime. Once the ligament surface approaches the particle, the subsequent breakup becomes self-similar and independent of the particle size. We derive an analytical expression for the pinch-off time based on the interplay between ligament stretching and Rayleigh-Plateau instability, which agrees quantitatively with simulations. Our results reveal a universal mechanism by which localized perturbations control the breakup of ligaments containing solid particles.
arXiv:2605.19147v1 Announce Type: new
Abstract: Large language models (LLMs) are highly susceptible to backdoor attacks (BAs), wherein training samples are poisoned using trigger-based harmful content. Furthermore, existing defenses have proven ineffective when extensively tested across BA patterns. To better combat BAs, we explore the use of LLM rewriting as a proactive defense against data poisoning. First, we theoretically show that when LLM rewriting utilizes open-book benign samples--termed open-book benign rewriting (OBBR)--the probability of a rewritten output being benign is strictly greater than that of closed-book rewriting. Thus, OBBR neutralizes harmful content by projecting training samples to the space of benign prompts. We then show that, in contrast to previous defenses, OBBR effectively mitigates a large number of existing BAs: across five known BAs and four widely used LLMs, OBBR increases safety performance by an average 51% compared to state-of-the-art BA defenses and 25.7% compared to closed-book rewriting methods. Finally, we show that OBBR is computationally efficient relative to other BA defenses, does not degrade model performance on natural language tasks after fine-tuning, and is capable of defending against non-trigger based data poisoning attacks.
arXiv:2605.19717v1 Announce Type: new
Abstract: Large Language Models (LLMs) can generate Computer-Aided Design (CAD), yet lack physical comprehension required for reliable engineering design. Instead of attempting to implicitly learn physical laws from data, we propose a Hybrid Agentic-Physical Architecture that embeds validated knowledge-based engineering tools directly into the decision making loop of autonomous AI agents. In this framework, engineering design is formulated as a closed-loop, sequential decision making process guided by explicit physical verification. Based on a load case, dedicated agents iteratively plan, generate, evaluate, and revise engineering designs using knowledge-based tools as a feedback signal. We introduce a benchmark dataset and metrics for assessing functional validity in generative CAD. Our system generates more complex and physically verified designs, with a 4.2 increase in structural complexity and improving compile rate by 3.5% compared to similar agentic methods. The codebase, prompts and dataset will be made publicly available to support reproducibility and future research.
arXiv:2605.19474v1 Announce Type: new
Abstract: We propose a discrete privacy mechanism exploiting beneficial properties of the novel privacy measure Pointwise Maximal Leakage (PML). Given the utility assignment characterized by every input-output letter pair, we study the mechanism design problem that satisfies PML privacy guarantees and maximizes the worst-case utility. Unlike popular privacy measures like Differential Privacy (DP), PML allows us to set some conditional probabilities in the mechanism to be zero and thereby preventing the occurrence of some low utilities while preserving a strict PML constraint. We show that the utility-safe mechanism, with low computational complexity, is optimal for the worst-case utility problem with an additional constraint on the output support set. We finally demonstrate the effectiveness in several numerical experiments. Due to DP's inability to have zeros in the mechanism, the design of privacy mechanisms that optimize the worst-case utility is underexplored, and this work shows that PML is a privacy measure that is perfectly suited for this purpose.
arXiv:2605.03230v2 Announce Type: replace
Abstract: SILMARILS is built from a minimal algebraic core over $\mathbb{F}_p$ using true randomness and perfect $2$-out-of-$2$ Shamir secret sharing. The framework supports both two-party and three-party modes. In the two-party setting, SILMARILS realizes a transferable designated-verifier (TDV) signature scheme. The designated verifier can simulate accepting transcripts indistinguishable from real ones, achieving Jakobsson-Sako-Impagliazzo DV security. The verifier may publish a receipt $r$ enabling public verification, yet even with $r$, no external party can tell whether a transcript was signed or simulated. As DV signatures permit simulation, standard EUF-CMA cannot hold for the designated verifier; instead, we prove $\mathsf{EUF\text{-}CMA}^{\neg\mathsf{DV}}$ security for all non-designated verifiers in both the random oracle model (ROM) and quantum random oracle model (QROM). In the three-party mode, adopting the broadcast model of Fitzi et al., we obtain a statistically secure signature protocol with simulation-based security and error $1/p$. We analyze security in the Pure IT model, the IT+ROM, and the QROM, extending the Fitzi et al. framework to quantum adversaries with classical I/O. Correctness, secrecy, transferability, and unforgeability for non-designated parties remain equivalent to simulation-based security. Thanks to its simple algebraic structure, SILMARILS offers very compact keys and signatures for the blockchain settings we target, where standardized PQC schemes are already more than sufficient. Our goal is not to compare SILMARILS with PQC, but to highlight its suitability for lightweight TDV authentication. A fair comparison with other DV schemes is omitted due to space and the complexity of aligning models.
arXiv:2605.19834v1 Announce Type: new
Abstract: To support operations and passenger-facing services, transit agencies need reliable passenger load trajectories. Currently, load estimates are typically inferred from imperfect sensing systems rather than fully observed, and the accuracy of modern automatic passenger counting (APC) systems still varies with station layout, flow intensity, and operating conditions. To address the challenges of robust passenger load estimation from heterogeneous data streams, including incremental count errors, evidence conflicts, and context-dependent sensor reliability, we propose a closed-loop, state-centric, multi-agent framework. This method enforces physical feasibility at every step, allocates trust dynamically among evidence sources, and feeds physics-derived violation residuals back into training for robustness improvement. The architecture consists of a unified stop-event backbone, a coupled Perception--Physical--Fusion loop for stop-by-stop inference, and optional trip-level macro-correction and closed-loop calibration modules.
arXiv:2605.19112v1 Announce Type: new
Abstract: Ordered logics and type systems have been used in a variety of applications
including computational linguistics, memory allocation, stream processing,
logical frameworks, parametricity, and enforcing security protocols. In most
formulations, ordered types are also linear, requiring each resource to be
used exactly once. Prior work by Kanovich et al. has investigated calculi
that relax this constraint through subexponentials within a linear ordered
logic. We generalize their work by using adjoint modalities to combine logics
with varying fine-grained structural properties, including weakening, left
contraction, right contraction, left mobility, and right mobility. We show
that the resulting sequent calculus admits cut elimination.
We further provide a natural deduction formulation in which structural rules
are implicit, and show that proof checking for this system is decidable. This
makes it a suitable foundation for an expressive adjoint programming language
or logical framework.
arXiv:2605.19824v1 Announce Type: new
Abstract: Recent attempts to support high-level scene interpretation and planning in Autonomous Vehicles (AVs) using ensembles of Large Language Models (LLMs) and Large Multimodal Models (LMMs) continue to treat time as a secondary property. This lack of temporal grounding leads to inconsistencies in reasoning about continuous actions, undermining both safety and interpretability. This work explores whether temporal conditioning within inter-agent communication can preserve or enhance coherence without introducing degradation in semantic or logical consistency. To investigate this, we introduce three planner architectures with progressively increasing temporal integration and evaluate them on curated subsets of the BDD-X dataset using semantic, syntactic, and logical metrics. Results show that while temporal conditioning reshapes reasoning style, it yields no statistically significant improvements in standard NLP-based correctness metrics. However, qualitative analysis reveals predictive hazard reasoning, stable corrective behavior, and strategic divergence in the Sentinel. These findings clarify the limits of prompt-based temporal grounding and establish the first empirical benchmark for temporal scene-to-plan reasoning.
arXiv:2605.18767v1 Announce Type: new
Abstract: Multi-hop question answering requires aggregating information from multiple documents, a critical capability for knowledge-intensive applications. A fundamental challenge lies in efficiently identifying the minimal relevant document set from retrieved candidates while maintaining high recall.
We present an efficient dual-view cascaded reranking framework for multi-hop document reranking. Operating as a lightweight post-retrieval stage over E5-base-v2 candidates, our architecture comprises: (1) a Local Scorer employing stacked cross-attention for fine-grained query-document relevance; and (2) a Global Scorer modeling inter-document dependencies via Transformer-based context aggregation. These views are dynamically fused through an Adaptive Gate conditioned on query semantics.
Under the fixed candidate set reranking setting with offline cached embeddings, our model achieves competitive results, particularly outstanding on MuSiQue with 99.4% Top-4 Recall and 97.8% Full Hit accuracy at 4.0 ms latency (249 QPS). It substantially outperforms 600M-parameter cross-encoders (BGE-Large: 92.0% Recall, Jina-v3: 90.1% Recall) while maintaining 5 to 6 times lower latency. Ablation studies validate that both Local and Global views contribute substantially to multi-hop performance.
arXiv:2604.16503v2 Announce Type: replace
Abstract: Training strong video generation models usually requires massive datasets, large parameter counts, and substantial compute. In this work, we ask whether strong text-to-video quality is possible at a much smaller budget: fewer than 10M clips and less than 100,000 H200 GPU hours. Our core claim is that part of the answer lies in how model capacity is organized, not only in how much of it is used. In video generation, prompt alignment, temporal consistency, and fine-detail recovery can interfere with one another when they are handled through the same pathway. Motif-Video 2B addresses this by separating these roles architecturally, rather than relying on scale alone. The model combines two key ideas. First, Shared Cross-Attention strengthens text control when video token sequences become long. Second, a three-part backbone separates early fusion, joint representation learning, and detail refinement. To make this design effective under a limited compute budget, we pair it with an efficient training recipe based on dynamic token routing and early-phase feature alignment to a frozen pretrained video encoder. Our analysis shows that later blocks develop clearer cross-frame attention structure than standard single-stream baselines. On VBench, Motif-Video~2B reaches 83.76\%, surpassing Wan2.1 14B while using 7$\times$ fewer parameters and substantially less training data. These results suggest that careful architectural specialization, combined with an efficiency-oriented training recipe, can narrow or exceed the quality gap typically associated with much larger video models.
arXiv:2605.16565v2 Announce Type: replace
Abstract: Skim is a speculative execution framework for web agents that exploits the predictable structure of purpose-built websites. Today's web-agent expense is not intrinsic to the tasks but a property of how agents are composed: frontier-model inference, browser rendering, and ReAct-style planning are applied to every step of every task regardless of complexity. Skim's key observation is that websites enforce stable URL patterns, answer formats, and task-to-trajectory mappings across queries of the same type, so most queries can bypass these heavyweight components entirely. An offline profiler captures these patterns once per site. At runtime, Skim matches each query to a template, synthesizes the destination URL, and extracts the answer with a small model. A lightweight verifier gates each fast-path output against the query and schema; rare misspeculations cascade to the full agent, warm-started by the fast path's final URL to preserve upstream trajectory progress. Across standard web-agent benchmarks paired with three backboneagents (WebVoyager, AgentOccam, BrowserUse), Skim reduces median per-task cost by 1.9x and latency by 33.4% with no accuracy loss.
arXiv:2605.19541v1 Announce Type: new
Abstract: In bandwidth-constrained communication such as satellite and underwater channels, speech must often be transmitted at ultra-low bitrates where intelligibility is the primary objective. At such extreme compression levels, codecs trained with acoustic reconstruction losses tend to allocate bits to perceptual detail, leading to substantial degradation in word error rate (WER). This paper proposes ClariCodec, a neural speech codec operating at 300 bit per second (bps) that reformulates quantisation as a stochastic policy, enabling reinforcement learning (RL)-based optimisation of intelligibility. Specifically, the encoder is fine-tuned using WER-driven rewards while the acoustic reconstruction pipeline remains frozen. Even without RL, ClariCodec achieves 4.64% WER on the LibriSpeech test-clean set at 300 bps, already competitive with codecs operating at higher bitrates. Further RL fine-tuning reduces WER to 3.55% on test-clean and 10.4% on test-other, corresponding to a 23% relative reduction while preserving perceptual quality.
arXiv:2601.22569v2 Announce Type: replace
Abstract: Large language model (LLM) based agents are increasingly used to automate financial transactions, yet their reliance on contextual reasoning exposes payment systems to prompt-driven manipulation. The Agent Payments Protocol (AP2) aims to secure agent-led purchases through cryptographically verifiable mandates, but its practical robustness remains underexplored. In this work, we perform an AI red-teaming evaluation of AP2 and identify vulnerabilities arising from indirect and direct prompt injection. We introduce two attack techniques, the Branded Whisper Attack and the Vault Whisper Attack which manipulate product ranking and extract sensitive user data. Using a functional AP2 based shopping agent built with Gemini-2.5-Flash and the Google ADK framework, we experimentally validate that simple adversarial prompts can reliably subvert agent behavior. Our findings reveal critical weaknesses in current agentic payment architectures and highlight the need for stronger isolation and defensive safeguards in LLM-mediated financial systems.
RefiningGPT: Specialized language Models for Automated Refinery Unit-level Process Diagram Synthesis
arXiv:2605.19704v1 Announce Type: new
Abstract: Applying LLMs to complex industrial processes remains challenging due to the semantic gap between natural language design intents and the rigorous physical logic of engineering. In the field of petroleum refining engineering, a critical bottleneck is the automated synthesis of Unit-level Process Diagrams (UPDs), which serve as the topological bridge connecting abstract requirements to concrete unit operations. In this paper, we propose RefineGPT, a domain-specialized agent for autonomous refinery design.RefineGPT adopts a hierarchical architecture in which a supervised fine-tuned small language model is responsible for selecting units that satisfy design requirements, while a large language model is used to connect these units to generate the final topology. To enable supervised training, we develop a pipeline that extracts latent process motifs from noisy, unstructured legacy topologies and synthesizes high-quality rationale-based Chain-of-Thought (CoT) training data. Empirical validation demonstrates that RefineGPT achieves substantial improvements in topological consistency and chemical engineering feasibility, establishing a high-fidelity pathway for AI-augmented industrial process synthesis.
arXiv:2604.02784v2 Announce Type: replace
Abstract: Vision-Language Models (VLMs) excel at multimodal tasks, but they remain vulnerable to hallucinations that are factually incorrect or ungrounded in the input image. Recent work suggests that hallucination detection using internal representations is more efficient and accurate than approaches that rely solely on model outputs. However, existing internal-representation-based methods typically rely on a single representation or detector, limiting their ability to capture diverse hallucination signals. In this paper, we propose EnsemHalDet, an ensemble-based hallucination detection framework that leverages multiple internal representations of VLMs, including attention outputs and hidden states. EnsemHalDet trains independent detectors for each representation and combines them through ensemble learning. Experimental results across multiple VQA datasets and VLMs show that EnsemHalDet consistently outperforms prior methods and single-detector models in terms of AUC. These results demonstrate that ensembling diverse internal signals significantly improves robustness in multimodal hallucination detection.
arXiv:2601.12006v2 Announce Type: replace
Abstract: In this paper we analyze the Finnish road network as a graph in order to measure whether the "rurality" or "urbanity" of an area correlates with local geometrical properties of the graph. Our primary motivation is the observation that the road systems in rural areas look similar to hyperbolic graphs, while in large cities they resemble more the Cayley graph of $\mathbb{Z}^2$. We do not aim for a comprehensive analysis, but rather wish to demonstrate that this observation can be measured and analyzed through looking at various "hyperbolicity measures" of randomly sampled geodesic triangles in the road graph.
arXiv:2605.19703v1 Announce Type: new
Abstract: Autonomous UAV flight in confined, wall-dense environments requires low-latency and reliable motion planning under strict safety constraints. Traditional optimization-based planners suffer from mapping latency and easily fall into local minima when navigating through dense structural obstacles. Meanwhile, existing end-to-end learning methods struggle to extract fine-grained geometric features from raw depth images and lack hard kinodynamic constraints, leading to unpredictable collisions near walls. To address these issues, we propose KIO-planner, an attention-guided single-stage trajectory planning framework. First, we integrate a Convolutional Block Attention Module (CBAM) into the perception backbone to adaptively focus on critical structural edges and traversable space. Second, we introduce a novel Dual Mapping mechanism--comprising physical bounds activation and a deterministic Geometric Safety Shield in the depth-pixel space--to enforce kinodynamic feasibility and collision-free flight without global map fusion. Extensive high-fidelity simulated experiments demonstrate that KIO-planner enables highly agile navigation at speeds up to 3.0 m/s. Compared to the state-of-the-art baseline, KIO-planner achieves lower inference latency (approximately 24 ms) and generates significantly smoother trajectories, reducing control cost by 28.4%. Most notably, our Dual Mapping substantially increases the worst-case safety margin, measured by minimum distance to obstacles, from 0.48 m to 0.76 m, ensuring fast, smooth, and safer navigation in highly constrained environments.
arXiv:2605.19702v1 Announce Type: new
Abstract: Color refinement is an important technique that works very well in practice for the graph isomorphism problem. Tinhofer graphs are the class of graphs for which refinement together with individualization correctly tests graph isomorphism against every other graph, irrespective of the choices of vertices made during individualization. Motivated by the fact that Tinhofer graphs form a natural boundary for efficient graph isomorphism tests based on color refinement, in this paper, we introduce a hierarchy of graph classes within the class of Tinhofer graphs. We call a graph $G$ $k$-Tinhofer if, after $k$ rounds of individualization and refinement, the resulting colored graphs remain isomorphic for every graph $H \cong G$, irrespective of the choices of vertices made during individualization.
Arvind et al. (2017) studied a hierarchy of graph classes motivated by color refinement - discrete, amenable, Tinhofer, and refinable graphs. We show that the $k$-Tinhofer hierarchy lies between the class of all graphs and Tinhofer graphs, with refinable graphs coinciding with the first level of the hierarchy. We obtain two characterizations of $k$-Tinhofer graphs: an algebraic characterization in terms of orbit partitions induced by pointwise stabilizers of automorphism groups, and a combinatorial characterization in terms of individualization-refinement trees and quotient graphs. For every fixed integer $k \ge 0$, there exist vertex-colored graphs that are $k$-Tinhofer but not $(k + 1)$-Tinhofer. For every fixed integer $k \ge 0$, the problem of deciding whether a given $k$-Tinhofer graph is ($k + 1$)-Tinhofer is $P$-hard under uniform $\mathsf{AC^0}$ many-one reductions. We show that testing isomorphism between an $(n - k)$-Tinhofer graph $G$ and an arbitrary graph $H$ is fixed-parameter tractable with respect to the parameter $k$.
arXiv:2605.18830v1 Announce Type: new
Abstract: Regression and Bayesian accounts of in-context learning (ICL) explain how demonstrations can induce predictors, while mechanistic analyses often identify compact activation directions that steer prompted behavior. However, it remains unclear whether structured demonstrations induce low-dimensional concept inference. We study this question through a concept-subspace view of ICL, in which tasks vary only along intrinsic concept coordinates, although inputs are observed in a high-dimensional ambient space. For ridge and least-squares ICL proxies, prediction decomposes exactly into concept-coordinate regression and off-subspace leakage. Under block-diagonal or near-block-diagonal covariance assumptions, the leading estimation and nuisance-sensitivity terms scale with the dimension of the concept subspace, while residual effects are controlled by cross-subspace coupling. This separation gives a mechanistic prediction: recoverable task information should concentrate in a low-dimensional, task-aligned activation subspace. On CounterFact-derived multi-relation prompts with Llama-3-8B, a 68--73-dimensional subspace of the 4096-dimensional residual stream restores 78.8% of the clean--corrupted accuracy gap, whereas patching the complementary subspace restores 0%. Concept swaps redirect predictions toward injected relations, while random and cross-task matched-rank controls are largely ineffective. Additional experiments on Qwen2.5-7B and a controlled cross-lingual rule task show the same qualitative pattern. These results support concept subspaces as compact, task-aligned mediators of recoverable ICL behavior in structured task families, without implying full-circuit recovery.
arXiv:2605.19700v1 Announce Type: new
Abstract: We present a novel dielectric terahertz-driven accelerator (DTA) that integrates a dual-pillar grating structure within a tapered parallel-plate waveguide (TPPWG). This compact setup enables efficient particle acceleration using multi-cycle, narrowband terahertz (THz) pulses. The TPPWG serves a dual role: it enhances the THz field via geometric tapering and delivers it to the dielectric structure by efficient coupling. Experimental validation of the THz field inside the waveguide is conducted using electro-optic sampling. Optimization of waveguide parameters through time-domain simulations reveals a sixfold peak electric field amplification at the end of the waveguide. The dielectric accelerator is tailored for maximum acceleration by adjusting the DTA pillar radius and vacuum channel gap for relativistic electron beams. Particle-in-cell (PIC) simulations demonstrate that the structure supports net acceleration with gradients up to 120 MeV per m for 0.1 GV per m field strengths, and can accommodate bunch charges up to 10 pC with minimal degradation. Energy spread evolution and beam dynamics are discussed in detail, including the role of phase slippage and bunch length. This work establishes the DTA-integrated TPPWG as a compact and scalable platform for high-gradient THz-driven acceleration, combining simple fabrication and design, strong field enhancement, and compatibility with existing electron sources. It opens new pathways toward practical, tabletop accelerators for scientific and industrial applications.
arXiv:2605.19338v1 Announce Type: new
Abstract: Frontier AI models and multi-agent systems have led to significant improvements in mathematical reasoning. However, for problems requiring extended, long-horizon reasoning, existing systems continue to suffer from fundamental reliability issues: hallucination accumulation, memory fragmentation, and imbalanced reasoning-tool trade-offs. In this paper, we introduce STAR-P\'olyaMath, a multi-agent framework that systematically addresses these challenges through meta-level supervision and structured Reasoner-Verifier interaction. STAR-P\'olyaMath is structured as an orchestrated state machine with nested challenge-step-replan loops, governed by a reasoning-free Python orchestrator that separates control from inference and bounds error propagation through trace-back and re-planning. Our key innovation is a persistent Meta-Strategist that maintains cross-attempt memory and exercises meta-level control by issuing high-level strategic guidance or mandatory directives, so the system can escape unproductive loops rather than stagnate or over-rely on tools. STAR-P\'olyaMath achieves state-of-the-art results on all eight top-tier competition benchmarks: AIME 2025-2026, MathArena Apex Shortlist, MathArena Apex 2025, Putnam 2025, IMO 2025, HMMT February 2026, and USAMO 2026. It obtains perfect scores on AIMEs, Putnam, and HMMT, and shows its largest margin on Apex 2025, scoring 93.75% compared with 80.21% by the strongest baseline GPT-5.5. Ablation studies show that the gains arise from the framework's orchestration rather than from model-level diversity since removing key components or substituting in mixed backbones consistently weakens performance. Code is available at https://github.com/Julius-Woo/STAR-PolyaMath.
arXiv:2512.07068v3 Announce Type: replace
Abstract: Uniform Meaning Representation (UMR) is a novel graph-based semantic representation which captures the core meaning of a text, with flexibility incorporated into the annotation schema such that the breadth of the world's languages can be annotated (including low-resource languages). While UMR shows promise in enabling language documentation, improving low-resource language technologies, and adding interpretability, the downstream applications of UMR can only be fully explored when text-to-UMR parsers enable the automatic large-scale production of accurate UMR graphs at test time. Prior work on text-to-UMR parsing is limited to date. In this paper, we introduce two methods for English text-to-UMR parsing, one of which fine-tunes existing parsers for Abstract Meaning Representation and the other, which leverages a converter from Universal Dependencies, using prior work as a baseline. Our best-performing model, which we call SETUP, achieves an AnCast score of 84 and a SMATCH++ score of 91, indicating substantial gains towards automatic UMR parsing.
arXiv:2605.19738v1 Announce Type: new
Abstract: Graph Anomaly Detection (GAD) aims to identify atypical graph entities, such as nodes, edges, or substructures, that deviate significantly from the majority. While existing text-rich approaches typically integrate structural context into the data representation pipeline using raw textual features, they often neglect the structural context of nodes. This limitation hinders their ability to detect sophisticated anomalies arising from inconsistencies between a node's inherent content and its topological role. To bridge this gap, we propose TERGAD (Structure-aware Text-enhanced Representations for Graph Anomaly Detection), A novel data augmentation framework that enriches structural semantics for GAD via the semantic reasoning capabilities of Large Language Models (LLMs). Specifically, TERGAD translates node-level topological properties into descriptive natural language narratives, which are subsequently processed by an LLM to derive high-level semantic embeddings. These embeddings are then adaptively fused with original node attributes through a gated dual-branch autoencoder to jointly reconstruct both graph structure and node features. The anomaly score is computed based on the integrated reconstruction error, effectively capturing deviations in both observable attributes and LLM-informed semantic expectations. Extensive experiments on six real-world datasets demonstrate that TERGAD consistently outperforms state-of-the-art baselines. Furthermore, our ablation studies validate the indispensable role of structural semantic guidance and the efficacy of the gated fusion mechanism. Code is available at https://github.com/Kantorakitty/TERGAD-main.
arXiv:2605.18892v1 Announce Type: new
Abstract: Federated learning (FL) enables collaborative learning of computer vision models, where privacy and regulatory constraints prevent centralizing data across devices or organizations. However, practical FL deployments often exhibit severe class imbalance and label skew, causing standard aggregation protocols to overfit dominant clients and degrade minority-class performance. We propose a data-free, class-wise contribution estimation and aggregation framework based on logit maximization (CELM) that does not require sharing raw data, client metadata, or auxiliary public datasets. The FL server probes client updates to obtain class-wise evidence scores and assembles a cross-client evidence matrix, which quantifies both per-class competence and class coverage. Using this matrix, we compute contribution weights that upweight clients providing strong, discriminative evidence for underrepresented classes. The resulting aggregation is stable due to simplex constraints and momentum smoothing, and it remains compatible with standard FL training pipelines. We evaluate the approach on representative vision benchmarks under controlled non-IID and pathological label splits, demonstrating that CELM-based aggregation improves robustness to imbalance and statistical heterogeneity, while yielding better performance without requiring any additional data exchange.