arXiv:2605.19357v1 Announce Type: new
Abstract: Large language models (LLMs) are increasingly applied to scientific research, yet existing evaluations often fail to reflect the fine-grained capabilities required in practice. Most benchmarks are manually curated or domain-generic, limiting scalability and alignment with real scientific use cases. In this paper, we propose a new framework named SciCustom to address the problem. It enables the custom construction of benchmarks from large-scale scientific data to evaluate application-specific scientific capabilities in LLMs. SciCustom first organizes scientific knowledge into ontology-grounded knowledge units with controlled granularity and trains a tagger to map large-scale data instances into this knowledge space. Given a custom requirement, relevant knowledge units are identified via voting-based multi-model consensus. These units enable relevance-aware benchmark retrieval via binary search, followed by proxy subset selection and data-grounded benchmark generation for efficient evaluation. Experiments in chemistry and healthcare demonstrate that SciCustom reveals fine-grained differences in LLM scientific capabilities that standard benchmarks overlook, while requiring neither expert annotation nor synthetic question generation. This work provides a scalable and application-aware foundation for benchmarking scientific capabilities in LLMs. The source code is available at https://github.com/yjwtheonly/SciCustom.
Science Journals
arXiv:2605.15768v2 Announce Type: replace
Abstract: Social simulation provides a compelling testbed for studying social intelligence, where agents interact through multi-turn dialogues under evolving contexts and strategically adapting opponents. Such environments are inherently non-stationary, requiring agents to dynamically adjust their strategies over time. However, most Large Language Model (LLM) based social agents rely on static personas, while existing approaches for enhancing social intelligence, such as offline reinforcement learning or external planners, are ill-suited to these settings, typically assuming stationarity and incurring substantial training overhead. To bridge this gap, we propose \textbf{ALSO} (\textbf{A}dversarial on\textbf{L}ine \textbf{S}trategy \textbf{O}ptimization), the first framework for online strategy optimization in multi-agent social simulation. ALSO advances social adaptation through two key contributions. (1) ALSO formulates multi-turn interaction as an adversarial bandit problem, where combinations of static personas and dynamic strategy instructions are treated as arms, providing a principled solution to non-stationarity without relying on environmental stability assumptions. (2) To predict rewards and generalize sparse feedback in multi-turn dialogues, ALSO introduces a lightweight neural surrogate to predict rewards from interaction histories, enabling sample-efficient exploration and continuous online adaptation. Experiments on the Sotopia benchmark demonstrate that ALSO consistently outperforms static baselines and existing optimization methods in dynamic environments, validating the effectiveness of adversarial online strategy optimization for building robust social agents.
arXiv:2605.19406v1 Announce Type: cross
Abstract: Time-based indoor positioning techniques rely on multiple access points (APs) and measurements between the user equipment (UE) and the APs. In dense indoor environments, occlusion-induced non-line-of-sight (NLoS) propagation introduces significant delays in these measurements, thereby degrading position estimation accuracy. To address this challenge, this paper proposes measurement selection strategies to improve position estimation accuracy. A ray-tracing (RT) simulator is employed to characterize the propagation environment and derive AP neighborhood information, which is subsequently used to design and evaluate different measurement selection strategies. The approaches explored include AP neighborhood-based cardinality selection, intersection and union of measurements from AP neighborhoods, and fixed measurement selection. Experiments demonstrate the efficacy of the proposed measurement selection strategies in environments under significant NLoS conditions.
arXiv:2605.19536v1 Announce Type: new
Abstract: In continuum topology optimization (TO), two essential procedures are involved: structural analysis through the solution of partial differential equations (PDEs) and the subsequent update of design variables. Both procedures can be addressed by training neural networks using the corresponding physical information. Accordingly, Physics-Informed Neural Network (PINN)-based algorithms have been developed for TO. However, PINN-based methods suffer from several notable limitations, including high computational cost, spectral bias, and limited adaptability in solving PDEs.To overcome these challenges, this study proposes a novel algorithm that incorporates two Higher-Order ReLU-based Kolmogorov-Arnold Networks (HRKANs). Specifically, a displacement-informed HRKAN (d-HRKAN) is designed to predict PDE solutions, while a sensitivity-informed HRKAN (s-HRKAN) is developed to perform sensitivity analysis for updating design variables. For convenience, the proposed approach is referred to as the Dual Physics-Informed Kolmogorov-Arnold Networks-based Topology Optimization (DPIKAN-TO) method. By leveraging learnable activation functions, the proposed neural networks can accurately approximate the responses of complex structural systems. Moreover, compared with conventional PINN-based methods, DPIKAN-TO demonstrates significantly improved computational efficiency and reduced computational cost. Numerical examples show that DPIKAN-TO can successfully identify optimal material layouts for linear structures, compliant mechanisms, and fluid-solid coupled systems. Furthermore, owing to the use of learnable activation functions, the proposed framework can be readily extended to structural optimization problems governed by new types of PDEs.
arXiv:2605.19394v1 Announce Type: new
Abstract: Adapting small instruction-tuned models to specialized domains often relies on supervised fine-tuning (SFT) on curated instruction-response examples, which is expensive to collect at scale. Synthetic training examples generated by a teacher LLM from a domain corpus can reduce this cost, but existing pipelines can produce homogenized outputs and do not consistently capture cross-passage or cross-document dependencies. We introduce EmbGen, a synthetic data generation pipeline that decomposes a corpus into entity-description pairs, reassembles them using semantic structure inferred from embedding similarity, and then generates question-answer (QA) pairs via proximity, intra-cluster, and inter-cluster sampling with cluster-specialized system prompts. We evaluate EmbGen against EntiGraph, InstructLab and Knowledge-Instruct on three datasets of varied semantic heterogeneity, under fixed token budgets (5 and 20 million tokens). We use lexical overlap metrics, an LLM-as-a-judge rubric, and Binary Accuracy, a composed metric combining Factual Accuracy and Completeness for evaluation. EmbGen improves Binary Accuracy on the most heterogeneous dataset by 12.5% at 5M and 88.9% at 20M tokens budget, relative to the strongest baseline, while remaining competitive across other datasets with lower heterogeneity.
arXiv:2603.06996v2 Announce Type: replace
Abstract: Adaptive Large Neighborhood Search (ALNS) is a prominent metaheuristic and a widely adopted approach for production and logistics optimization. However, it has long relied on hand-crafted components built on expert experience, which makes development slow and costly to adapt to new problems. This paper proposes a closed-loop, large-language-model-driven evolutionary framework that decouples ALNS and automatically rebuilds all of its components. We break ALNS into seven key modules: destroy, repair, operator selection, weight update, initial solution construction, acceptance rule, and destroy-rate control, and evolve each module through a dedicated task. By incorporating the Multi-dimensional Archive of Phenotypic Elites mechanism, the framework maintains a multi-dimensional elite archive to simultaneously drive the evolution of solution quality and strategic diversity. In addition, we design multiple mechanisms, including parallel and sequential multi-module evolution as well as single-expert-driven and multi-expert-driven evolution, to systematically evaluate the impact of different evolutionary paradigms on algorithm generation performance. Evaluations on Traveling Salesman Problem and Capacitated Vehicle Routing Problem benchmarks demonstrate that evolved algorithms consistently outperform optimized classic ALNS baselines under both fixed-iteration and fixed-time limits. The framework also shows a degree of generalizability and cross-problem transferability. Code analysis also uncovers several counterintuitive yet meaningful design patterns that emerged naturally during evolution, offering practical and theoretical insights for future ALNS design. Finally, comparisons across multiple language models highlight clear differences in their ability to support evolutionary algorithm design, helping guide model selection for real-world engineering use.
arXiv:2605.19826v1 Announce Type: new
Abstract: Operators of safety-critical industrial processes increasingly rely on digital twins to screen control interventions, but such simulators rarely carry certified safety guarantees. Wastewater treatment plants exemplify the gap: operators face a daily safety-efficiency trade-off where aerating too little risks effluent violations and nitrous-oxide (N2O) spikes, and aerating too much wastes energy. We develop an explainable digital twin for aeration and dosing setpoints. CCSS-IX, the simulator, is a bank of interpretable locally linear state-space "experts" adaptively mixed by a context-aware gating network, building on a continuous-time regime-switching scaffold. A runtime decision layer applies conformal risk control to abstain, reopen, or return a falsifying temporal witness for any operator-proposed action that cannot be statistically certified. The artificial-intelligence contribution is twofold: an identifiable, context-conditioned structured surrogate that retains operator-readable dynamics, and a self-falsifying decision rule with finite-sample coverage guarantees. The engineering contribution is a validated, end-to-end decision-support pipeline, tested on a 1000-step slice of the Aved{\o}re full-scale plant (42.6% sensor missingness, 2-minute sampling), the Agtrup/BlueKolding full-scale plant in Denmark, and the Benchmark Simulation Model No. 2 (BSM2) international benchmark, under a matched ten-seed protocol. The static structured ensemble lies within 0.78% root-mean-square error of an unconstrained black-box reference, and the adaptive variant within 1.08%. The calibrated reopen rule cuts aggregate two-plant regret by 43.6% at an unsafe-action cost weight of 4 and eliminates unsafe chosen actions on the BSM2 main slice. Event-aligned temporal witnesses prevent 93 of 187 false-safe N2O approvals, about 4.65x the dyadic baseline (paired McNemar p < 1e-21).
arXiv:2605.20180v1 Announce Type: cross
Abstract: One central theme in quantum photonics is tailoring the interactions between atoms/spins and their electromagnetic (EM) environments. Considerable effort has focused on engineering spontaneous emission by shaping EM environments, known as the Purcell effect. However, photonic environment control of pure dephasing, which is a complementary paradigm of non-unitary atom/spin couplings with EM environments, remains largely unexplored. Here, we introduce a nanophotonic approach to modify qubit pure dephasing dynamics. Unlike Purcell engineering that tailors photonic environments at qubit resonance frequencies (typically optical/near-infrared), we develop ultra-subwavelength spin noise metasurfaces for efficient broadband control of low-frequency (e.g., $\sim$MHz) photonic environments far off-resonant with atoms/spins for dephasing engineering. We experimentally demonstrate our approach using lithographically defined CoFeB metasurfaces and shallow nitrogen-vacancy (NV) centers in diamond. Instead of modified spontaneous emission, we observe modified NV pure dephasing dynamics near different spin noise metasurfaces. We further isolate metasurface-controlled dephasing from other dephasing mechanisms (e.g., spin bath) by measuring the NV ensemble dephasing noise spectrum with dynamical decoupling spectral decomposition techniques. Our results establish a new frontier in engineering quantum light-matter interactions with nanophotonic structures.
arXiv:2603.28524v2 Announce Type: replace-cross
Abstract: An accurate and efficient numerical electromagnetic model for superconducting qubits is essential for characterizing and minimizing design-dependent dielectric losses. The energy participation ratio (EPR) is the commonly adopted metric used to evaluate these losses, but its calculation presents a severe multiscale computational challenge. Conventional finite element method (FEM) requires 3D volumetric meshing, leading to prohibitive computational costs and memory requirements when attempting to capture singular electric fields at nanometer-thin material interfaces. To address this bottleneck, we propose SesQ, a surface integral equation simulator tailored for the precise simulation of the EPR. By applying discretization on 2D surfaces, deriving a semi-analytical multilayer Green's function, and employing a dedicated non-conformal boundary mesh refinement scheme, SesQ accurately resolves singular edge fields without an explosive growth in the number of unknowns. Validations with analytically solvable models demonstrate that SesQ accelerates capacitance extraction by roughly two orders of magnitude compared to commercial FEM tools. While achieving comparable accuracy for capacitance extraction, SesQ delivers superior precision for EPR calculation. Simulations of practical transmon qubits further reveal that FEM approaches tend to significantly underestimate the EPR. Finally, the high efficiency of SesQ enables rapid iteration in the layout optimization, as demonstrated by minimizing the EPR of the qubit pattern, establishing the simulator as a powerful tool for the automated design of low-loss superconducting quantum circuits.
arXiv:2002.09053v3 Announce Type: replace
Abstract: Pedestrian detection benefits from deep learning technology and gains rapid development in recent years. Most of detectors follow general object detection frame, i.e. default boxes and two-stage process. Recently, anchor-free and one-stage detectors have been introduced into this area. However, their accuracies are unsatisfactory. Therefore, in order to enjoy the simplicity of anchor-free detectors and the accuracy of two-stage ones simultaneously, we propose some adaptations based on a detector, Center and Scale Prediction(CSP). The main contributions of our paper are: (1) We improve the robustness of CSP and make it easier to train. (2) We propose a novel method to predict width, namely compressing width. (3) We achieve the second best performance on CityPersons benchmark, i.e. 9.3% log-average miss rate(MR) on reasonable set, 8.7% MR on partial set and 5.6% MR on bare set, which shows an anchor-free and one-stage detector can still have high accuracy. (4) We explore some capabilities of Switchable Normalization which are not mentioned in its original paper. The code is publicly available at https://github.com/WangWenhao0716/Adapted-Center-and-Scale-Prediction.
arXiv:2605.19427v1 Announce Type: cross
Abstract: A conservation-consistent boundary condition is proposed for nonlinear models of soluble-surfactant-laden falling films, ensuring exact conservation of total surfactant mass. The formulation resolves an inconsistency in widely used reduced models, Pascal et al. (PRF, 2019), D'Alessio et al. (JFM, 2020), which exhibit a gradual drift of mass during nonlinear evolution in a closed periodic domain. We show that this originates from an inconsistency in the surface transport reduction and derive a corrected boundary condition that removes this defect. As the discrepancy appears only at the nonlinear order, linear stability results remain unaffected, explaining why the issue has remained unnoticed.
arXiv:2605.20033v1 Announce Type: new
Abstract: Multimodal large language models often generate reasoning chains containing subtle errors that lead to incorrect answers. Current verification approaches have notable limitations. Learned critics need extensive labeled data and show inconsistent performance across different tasks. Meanwhile, existing training-free methods simply average scores from different sources, missing a key insight: when these scores disagree, that disagreement itself carries important information about whether a reasoning step is truly valid or not. We propose a training-free verification approach that treats step-wise verification as a coordination problem among specialized judges. We formalize these judges' interaction as a Nash equilibrium game where agreement signals valid steps while disagreement reveals instability. Our method computes equilibrium scores through a closed-form solution, enabling both disagreement-aware filtering and stability-conscious ranking of reasoning steps. Evaluated across six benchmarks, our approach achieves consistent improvements of 2.4% to 5.2% over baseline models and shows competitive performance against learned critics, demonstrating that cross-modal agreement (not just average confidence) provides robust verification signals without task-specific adaptation.
arXiv:2605.05569v3 Announce Type: replace-cross
Abstract: This paper shows that the semi-dual formulation of the optimal transport problem has a degenerate saddle-point structure, and that its numerical solution is equivalent to solving a constrained optimization problem. We derive necessary and sufficient conditions for the convergence of Monge maps without requiring optimality of the dual potential. This analysis helps explain why, in practice, numerical algorithms often require more iterations to update the transport map than the potential.
arXiv:2605.19915v1 Announce Type: new
Abstract: Classical models of opinion dynamics assume human participants with bounded rationality and limited coordination. The rise of LLM-based agents introduces a qualitative shift: agents can now participate in online discussions at scale, maintain consistent persuasion strategies, and coordinate systematically. This paper argues that LLM agents make collective belief dynamics programmable, enabling deliberate steering of population-level beliefs. We term this emerging problem programmable collective belief control. Through controlled multi-agent simulations, we provide proof-of-concept evidence that coordinated AI agents can induce measurable belief shifts that stabilize within a few interaction rounds. We identify four structural properties (indistinguishability, persistence, contextuality, and configurability) that make detection and defense fundamentally difficult. Based on these findings, we outline a research agenda spanning theoretical foundations for adversarial belief dynamics, operational methods for system-level detection and intervention, and simulation infrastructure for scalable experimentation. Our goal is not to present a complete solution, but to articulate why this problem demands urgent attention and to provide a conceptual foundation for future work.
arXiv:2605.19587v1 Announce Type: new
Abstract: Indoor scene synthesis underpins embodied AI, robotic manipulation, and simulation-based policy evaluation, where a useful scene must specify not only what the environment looks like, but also how its objects are structured. Existing pipelines, however, typically represent generated content as static meshes and inherit articulation only from curated asset libraries, which limits object-level controllability and prevents new interactable assets from being produced on demand. We address this gap by formulating physically interactable indoor scene synthesis as programmatic world generation, and present SceneCode, a framework that compiles a natural language prompt into an executable, code-driven indoor world rather than a collection of opaque meshes. A room-level agentic backbone first turns the prompt into a structured house layout and emits per-object AssetRequests through a planner--designer--critic loop. Each request is then routed to one of five code-generation strategies and converted into a synthesized part-wise Blender Python programs that are validated through an execution-guided repair-and-refine loop. The resulting programs are compiled into simulation-ready assets, and exported as SDF for physics simulation. A persistent scene-state registry links object requests, executable programs, rendered geometry, and simulation assets, turning scene assembly into a traceable and locally editable world-building process. We evaluate SceneCode across scene-level synthesis, object-level asset quality, human judgment, and downstream robot interaction. Results show that executable world programs improve prompt-faithful indoor scene generation and produce assets with cleaner mesh structure, and simulator-loadable articulation metadata. Project page: https://scene-code.github.io/.
arXiv:2605.19910v1 Announce Type: new
Abstract: The simulation of quantum transport in nanodevices requires the solution of the Dyson and Keldysh equations, a task dominated by the inversion of massive, block-tridiagonal matrices. While the Recursive Green's Function (RGF) method has long been the standard $O(N)$ solver for quasi-1D systems, its formulation has typically been restricted to sequential execution and nearest-neighbor interactions. In this work, we carefully reformulate RGF through the lens of Domain Decomposition and Schur Complement theory. This allows us to extend the recursive formalism to block $n$-diagonal systems (handling higher-order stencils) and to derive a parallel algorithm, Domain-Decomposition based RGF (DDRGF), which stitches macroscopic domains via reduced interface systems. We explore data dependencies in DDRGF in detail, by means of block-sparse structures and tracing back to the desired output as a block tridiagonal approximation, giving a clear, reproducible and extensible formulation. We validate these algorithms using \texttt{LibNEGF.jl}, a Julia-based implementation, demonstrating that the structural insights of domain decomposition provide a robust pathway for high-performance quantum transport simulations on modern multi-core clusters. The theory presented here lays down the base for tackling the Keldysh problem, to be similarly handled in future stages of our work. Although the target here is the acceleration of kernels in the non-equilibrium Green's function method, the algorithms and the implementations presented can be immediately used in any application involving block $n$-diagonal systems.
arXiv:2605.19928v1 Announce Type: new
Abstract: Counterfactual Regret Minimization (CFR) is the dominant algorithmic family for solving large imperfect-information games, underpinning breakthroughs such as Libratus and Pluribus in No-Limit Texas Hold'em poker. In real-time game-playing systems, the solver must compute a near-equilibrium strategy within a strict time budget of only a few seconds per decision, and the number of CFR iterations completed in this window directly determines play strength. We present \textbf{Parallel CFR}, the first parallelization framework for real-time depth-limited CFR solving that seamlessly integrates pruning, abstraction, and advanced CFR variants. We decompose each CFR iteration into a pipeline of seven stages and identify two orthogonal dimensions of parallelism: \emph{by information set} and \emph{by tree node}. Leaf node evaluation is offloaded to GPUs via batched neural network inference, creating a heterogeneous CPU--GPU pipeline. Experiments on Heads-Up No-Limit Texas Hold'em demonstrate that Parallel CFR achieves $3.3$--$3.4\times$ speedup over the single-threaded baseline on postflop streets, with per-iteration time of ${\sim}47$--$54$~ms on a depth-limited game tree with over $1$ billion histories. All experiments run on a single desktop-class device (NVIDIA DGX Spark), enabling hundreds of CFR iterations within a typical real-time decision budget without requiring datacenter-scale infrastructure.
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:2605.19622v1 Announce Type: new
Abstract: Representation learning with Vision Transformers (ViTs) has advanced rapidly, yet the utility of large-scale models in spatially sensitive tasks is hindered by spurious tokens. Prior efforts to mitigate this have been limited, often defining these artifacts narrowly, for example, as simple high-norm outliers. We argue that this scope is insufficient. For dense prediction tasks, we posit that any token failing to encode location-aligned semantics should be treated as a spurious artifact. This broader definition reveals a more complex problem, leading us to systematically categorize and characterize three fundamental types of spurious tokens that corrupt spatial representations. Based on this comprehensive diagnosis, we propose UniRefiner, a universal refinement framework that teaches pre-trained ViTs to self-dispose of these artifacts. UniRefiner uses contrastive registers to explicitly isolate and redistribute spurious tokens via a dual objective: (i) it aligns image tokens with filtered regular tokens to preserve semantics, and (ii) it aligns register tokens with detected spurious tokens to capture the spurious signals. Our method requires only a few epochs of fine-tuning on ~5k images to refine diverse ViTs, including massive models like EVA-CLIP-8B and InternViT-6B. Experiments demonstrate consistent and significant improvements: notably, the refined EVA-CLIP-8B achieves 51.9\% mIoU on ADE20K (+9.4\%), surpassing specialized vision models like DINOv2 (49.1\%), while zero-shot segmentation accuracy improves by up to 22\%. UniRefiner unlocks the latent spatial potential of existing large-scale foundation models, paving the way for their broader application.
arXiv:2306.12935v4 Announce Type: replace
Abstract: The asynchronous and unidirectional communication model supported by mailboxes is a key reason for the success of actor languages like Erlang and Elixir for implementing reliable and scalable distributed systems. Although actors eliminate many of the issues stemming from shared memory concurrency, they remain vulnerable to communication errors such as protocol violations and deadlocks. Behavioural types make it possible to detect communication errors early in the development process, but most work has addressed channel-based languages rather than actor languages.
Mailbox types are a novel behavioural type system for actors first introduced for a process calculus, and capture the contents of a mailbox as a commutative regular expression. Due to aliasing and nested evaluation contexts, moving from a process calculus to a programming language is challenging. This paper presents Pat, the first programming language incorporating mailbox types, and describes an algorithmic type system. Pat is a higher-order functional language with sums, products, and lists, along with interfaces that allow finer-grained reasoning about mailbox contents. Typechecking in Pat detects four classes of behavioural error: protocol violation, unexpected message, forgotten reply and self-deadlock, as well as the usual data type errors.
The Pat type system makes essential use of quasi-linear typing to tame some of the complexity introduced by aliasing. We make use of a co-contextual algorithmic type system, achieved through a novel use of backwards bidirectional typing, and we prove it sound and complete with respect to our declarative type system. We implement a type checker, and use it to demonstrate the expressiveness of Pat on a factory automation case study and a series of examples from the Savina actor benchmark suite. This establishes a foundation for applying mailbox typing to practical actor languages.
arXiv:2605.20072v1 Announce Type: new
Abstract: Large Language Models are increasingly proposed as cognitive components for robotic systems, yet their opaque decision processes make it difficult to explain success or failure in closed-loop embodied tasks. Following an empirical AI methodology, we study embodied LLM agents behaviorally by varying the information available to the agent and measuring the resulting changes in behavior. Using the Lockbox, a sequential mechanical puzzle with hidden interdependencies, we evaluate LLMs across RGB, RGB-D, and ground-truth symbolic observations in a physical robotic setup and use controlled simulation to probe the resulting behavior. Counterintuitively, agents perform best under raw RGB input and worst under perfect ground-truth observations. In simulation, we probe this effect by randomly flipping perceived action outcomes and find that moderate noise improves performance, peaking at a 40% flip probability with a 2.85-fold success rate increase over the noise-free baseline. Further analysis links this gain to a reduction in repetitive action loops. These findings suggest that success rates alone are insufficient for evaluating LLMs, as measured performance may reflect the interaction between perceptual errors and reasoning failures rather than robust problem solving.
arXiv:2605.18930v1 Announce Type: new
Abstract: Memory-augmented large language model (LLM) agents use iterative reflection and self-evolution to solve complex tasks, but these mechanisms introduce security risks. Existing agentic memory attacks require privileged access or explicit malicious content, making them detectable by advanced safety filters. This leaves a subtler attack surface underexplored: whether adversaries can induce agent to generate experiences that appear locally correct and semantically plausible yet induce harmful generalization during reflection. We find that reflective agents are vulnerable to such clean experiences, especially when paired with severe but plausible hypothetical consequences. Based on this observation, we introduce Obsessive Experience Poisoning (OEP), a low-privilege black-box attack requiring no direct control over the system prompt or memory database. OEP constructs adversarial clean edge-cases that combine locally correct solutions, non-transferable methods, and severe consequences, biasing reflection toward risk-averse rule formation. During memory consolidation, agents may over-trust self-generated reflections and distill localized experiences into high-priority but over-generalized rules, causing downstream failures. Evaluations across three domains show that OEP achieves ASR above 50\% with GPT-4o agents, and outperforms existing attacks under LLM auditing defense.
arXiv:2605.20090v1 Announce Type: new
Abstract: Multi-modal remote sensing images are vital for Earth observation, yet complete paired observations are often scarce in practice. Existing generative methods commonly address this problem through isolated pairwise modality translation, but their versatility and scalability remain limited as the number of modalities and generation tasks increases. Here, we develop a generative foundation model MetaEarth-MM for multi-modal remote sensing imagery, enabling paired joint generation and any-to-any translation across five modalities within a unified model. Recognizing the intrinsic scene consistency underlying multi-modal observations, we introduce a scene-centered joint modeling paradigm in MetaEarth-MM. Unlike previous methods that rely on direct appearance-level cross-modal mapping, our model organizes the generation around the underlying scene content. Specifically, MetaEarth-MM adopts a decoupled architecture that first infers a latent scene representation from available observations, and then generates target modalities conditioned on this intermediate state. To support training, we further construct EarthMM, a large-scale dataset comprising 2.8 million multi-resolution global images with 2.2 million aligned pairs. Extensive experiments demonstrate that MetaEarth-MM not only exhibits strong generative capability and robust generalization across diverse generation tasks, but also supports downstream tasks at both data and representation levels, highlighting its potential as a general foundation model for cross-modal Earth observation. The code and dataset will be available at https://github.com/YZPioneer/MetaEarth-MM.
arXiv:2605.19986v1 Announce Type: new
Abstract: Fine-grained manipulation marks a regime where global scene context no longer suffices, and success hinges on the tight coupling of local attribute grounding, high-fidelity spatial perception, and constraint-respecting motor execution. However, current embodied AI benchmarks collapse these capacities into binary success rates, systematically inflating reported capabilities by up to 70% and masking the architectural bottlenecks that impede real-world deployment. We introduce MetaFine, a diagnostic meta-evaluation framework that disentangles manipulation competency along three axes: understanding, perception, and controlled behavior. Built on a compositional task graph, MetaFine absorbs heterogeneous external benchmarks and reconstructs them into diagnostic scenarios of varying complexity under a unified protocol. Evaluating state-of-the-art vision-language-action (VLA) models through this lens exposes severe dimension-specific failures invisible to conventional metrics. Through targeted causal intervention, we identify the visual encoder's ability to preserve local spatial structure as a key bottleneck for fine-grained precision: improving it directly unlocks previously inaccessible manipulation capabilities without modifying downstream policies. MetaFine further supports hybrid real-sim validation, using limited paired real-world rollouts to calibrate scalable simulation-based estimates for more stable physical benchmarking. By shifting evaluation from ranking to diagnosis, MetaFine turns benchmarking into an actionable compass for repairing the layered capacities underlying genuine physical dexterity. The MetaFine framework, benchmarks, and supporting resources will be publicly released at our project page: https://metafine.github.io/.
arXiv:2605.20110v1 Announce Type: new
Abstract: Referring segmentation grounds natural-language queries to pixel-level masks, but extending it to complex scenarios with multiple instances, cross-category groups, or open-ended target sets remains challenging. Previous Large Vision Language Model (LVLM)-based methods represent referred targets with one or more special tokens sequentially, treating multiple targets as separate outputs rather than a coherent set and offering little incentive to capture set-level properties such as completeness and mutual exclusivity. We reformulate open-ended referring segmentation as explicit set-level concept prediction and propose Set-Concept Segmentation (SetCon), which uses LVLM-generated natural-language concepts, instead of segmentation-specific tokens, as semantic conditions for joint mask-set decoding. A hierarchical semantic decomposition first predicts a shared set-level concept defining the target scope and then refines it into fine-grained concept groups aligned with target subsets. To support this, a two-stage annotation pipeline augments existing reasoning segmentation datasets with hierarchical semantic supervision (236k samples, 784k concept phrases). SetCon achieves state-of-the-art results on image benchmarks (+3.3 gIoU on gRefCOCO, +12.1 gIoU on MUSE), with margins that grow as the number of referred targets increases. The concept interface also transfers to video under a detect-and-track setting, yielding new state-of-the-art results on seven referring video benchmarks, including +10.9 J&F on MeViS and +12.4 J&F on Ref-SeCVOS.