Papers

Agricultural systems make decisions over heterogeneous entity types, textual judgments, and knowledge-question options. Generating a textual answer adds decoding and extraction steps to decisions that ultimately require a candidate choice. We introduce AgriJev, a family of adapted Qwen3.5 models at 0.8B, 2B, 4B, and 9B. A shared interface conditions on agricultural text, a judgment question, and candidate definitions, returning scores and relative probabilities in one forward call. We prepare source-aware training records and frozen evaluations from existing agricultural resources, retaining original groups and annotation provenance. Comparisons include original Qwen3.5 models and matched agricultural SFT models, both evaluated through native answer generation. Across 17 external test views, AgriJev-9B reaches 75.44% mean accuracy, exceeding the original and SFT generators by 5.89 and 3.79 percentage points. It outperforms the original generator on 13 views and narrows the mean gap to hosted Jev 1.13.0. On a fixed A100 workload, its median scoring latency is approximately half that of same-scale answer generation. Task-dependent gains and failures characterize a practical choice among agricultural model weights and inference contracts.

Post-training often improves the performance of large audio-language models on audio understanding and reasoning benchmarks. However, these improvements may reflect adaptation to specific question formats, such as multiple-choice questions, rather than genuine advances in audio understanding and reasoning capabilities. To investigate what LALMs actually learn from post-training, we conduct controlled experiments that disentangle improvements in audio understanding from adaptation to specific question formats. Across different models and post-training methods, we find that much of the reported improvement persists even without informative audio, but largely disappears when the question format changes. These findings suggest that post-training often teaches models to exploit task-specific textual cues and answer patterns rather than developing generalizable audio understanding and reasoning capabilities.

Multimodal in-context learning (ICL) is an important capability of multimodal large language models (MLLMs). Its performance is closely tied to the demonstrations it uses, making demonstration retrieval a core problem for ICL. Common methods use fixed scoring methods to retrieve all the demonstrations at once, which may introduce redundancy and noise. Context-aware retrieval methods have emerged to address this problem by allowing incremental retrieval from a growing context. However, none of them lets the MLLM learn how to judge what new demonstrations it needs and then carry out progressive retrieval on its own. To this end, we introduce ActICL, which trains an MLLM agent to acquire and use an ICL context via tool-augmented reasoning. An ActICL agent analyzes its own information gaps during iterative reasoning and tool-use actions. Based on these gaps, it performs targeted retrieval to acquire a set of initial candidates. It can then use the existing context to further filter the candidates and refine the context, or it can iteratively perform a new retrieval. When the agent judges that the context is sufficient for the task, or when the resource limit is reached with the user request met, it sends the acquired context to an answering model, either the agent itself or another MLLM. We use a two-stage post-training procedure that jointly optimizes the agent's acquisition and answering. With an adaptive number of demonstrations, ActICL improves average performance by 11.6 percentage points over standard similarity-based retrieval across seven benchmarks and three MLLMs, and by 4.4 points compared to the strongest prior baseline. The contexts acquired by ActICL agents can also be used effectively by frontier models. When GPT-5.6-Sol performs 4-shot ICL, ActICL's contexts improve its performance by 3.5 points over standard retrieval, averaged over three benchmarks.

Questions about an agent's history often require aggregating events, which retrieval alone does not resolve. We combine source-preserving evidence access with executable aggregation. Trajectory Memory (TM) retains the original trajectory and constructs a source-addressable reading view at question time. TallyMem augments it with a reusable event ledger, model-generated event specifications and deterministic execution. With a commit-specific adaptation on 776 questions from 99 held-out MINTEval repositories, TallyMem raises Qwen3-32B exact match from 35.2% to 49.0%, a 13.8-point gain over TM; gains span 12.8-13.9 points across three readers. A controlled probe with event sets parsed independently from raw logs shows an 11.0-point gain, from 82.8% to 93.8%, without routing. On full AMA-Bench, the routed system yields a smaller 1.3-point gain, from 69.8% to 71.1%. These results support executable aggregation on tasks that require combining events, with benefits dependent on the task and event specification, the largest residual system-error category in our diagnostic analysis.

Long-term memory remains a fundamental bottleneck for large language model (LLM) agents across sessions. Existing systems typically adopt a write-then-retrieve paradigm, storing context-dependent dialogue fragments and retrieving them via nearest-neighbor search in a continuous embedding space. This design yields unstable access units and systematically misses logically related but lexically distant evidence. We propose MemCode, a long-term agent memory framework that couples retrieval-oriented writing with discrete associative retrieval. MemCode rewrites dialogue into self-contained atomic memory units (AMUs), maps AMUs into a learnable codebook with isomorphic semantic quantization (ISQ), and uses codebook co-occurrence topology as a retrieval path complementary to vector search. For long-term memory stores, MemCode separates online atomic writes from asynchronous offline consolidation. On LongMemEval-S and LoCoMo with GPT-4o-mini and Qwen3-30B-A3B-Instruct-2507 backbones, MemCode achieves the best overall accuracy while keeping per-dialogue memory construction within roughly 40~130 seconds, making it 3~21 times faster than LightMem and over an order of magnitude faster than A-MEM and MemoryOS.

Chest X-Ray (CXR) is a primary modality for diagnosing pulmonary diseases, yet the overlap of bony structures often obscures critical radiographic findings, increasing the risk of diagnostic uncertainty. While Dual-Energy Subtraction (DES) is the clinical benchmark for bone suppression, its widespread adoption is primarily hindered by specialized hardware requirements and increased radiation exposure. Recently, diffusion-based generative models have shown promise for this task, yet they are still constrained by reconstruction-biased latent compression, prior-induced domain gap, and a scarcity of high-quality paired datasets. To address these challenges, we propose DeBoneDiT, a depth-driven conditional bridge diffusion Transformer architecture for efficient high-resolution bone suppression. We first introduce Depth Auto-Encoder (DAE), a structure-aware vision tokenizer for latent compression. By incorporating a multi-level depth feature loss derived from the pretrained Depth Anything V2 encoder, DAE preserves both visual fidelity and structural consistency. Second, we formulate bone suppression as a source-conditioned Brownian bridge diffusion process, leveraging CXRs as both the diffusion prior and structural guidance to narrow the source-to-target domain gap and maintain anatomical plausibility. Furthermore, we have constructed and released SZCH-X-Rays, the largest publicly available high-quality paired dataset for bone suppression to date, comprising 741 pairs of CXR and DES soft tissue images. Extensive experiments and downstream evaluation demonstrate that DeBoneDiT achieves superior bone suppression performance with reduced computational overhead.

In initialized spacecraft tracking, a short-horizon motion prediction remains available as event support changes. Observation-Aligned Event Innovation (OA-EI) uses events to correct selected components of this prediction. A constant-body-twist model propagates the complete pose; the event encoder proposes rotational and normalized image-plane translation corrections, retaining propagated axial depth. Candidate magnitudes come from a bounded Scale Gate and a deterministic activity factor, with empty windows returning the prediction unchanged. Evaluation uses four FRESH physical-event trajectories, twenty rollout starts, and three training seeds. Each rollout starts from two reference poses and then uses self-estimates. At 7.7 seconds, rotation and translation errors fall from 41.63{\deg}/57.28 mm for prediction alone to 15.73{\deg}/22.41 mm for OA-EI, reductions of 62.2% and 60.9%. Failure rates fall from 20.0% to 1.67%. Matched ablations favor the rotation-bearing interface for overall pose accuracy; combining scaling with activity support gives the lowest errors at every tested event-retention level. On Jetson Orin Nano, the measured in-memory update takes 44.56 milliseconds at the median and 58.77 milliseconds at P95.

Emerging data-intensive space applications generate large volumes of data that must be processed, stored, replicated, and delivered efficiently. However, relying on terrestrial cloud regions can expose satellite data to high delivery latency and loss of timeliness due to intermittent satellite-to-ground visibility and limited access to ground infrastructure. Space Data Centers (SDCs), which organize collaborative satellites into logical compute and storage clusters, offer a promising approach for in-orbit data management and service continuity. In this paper, we first evaluate how satellite clustering with different sizes affects satellite visibility and gateway outage. We then propose BIER-DC, a stateless one-to-many communication protocol tailored for satellite-based SDCs. Unlike terrestrial multicast protocols designed for stable data center fabrics, BIER-DC decouples inter-SDC forwarding from intra-SDC delivery and encodes compact cluster-level forwarding information in packet headers. Comparative evaluations against BIER, BIER-TE, and Yeti show that BIER-DC significantly reduces packet-header length and byte-hop overhead, improving effective goodput and scalability for next-generation LEO space data centers.

We study learning-augmented streaming algorithms for estimating the value of Boolean Maximum Constraint Satisfaction Problems (Max-CSPs). Specifically, we consider streaming algorithms equipped with an $\varepsilon$-accurate oracle: for each variable, the oracle outputs a label in $\{-1,1\}$ that agrees with a fixed optimal assignment with probability $1/2+\varepsilon$, and disagrees otherwise. Prior work by Dong, Peng, and Vakilian [ITCS 2025] showed that, for the special case of Max-Cut, such an oracle enables a $(1/2+\Omega(\varepsilon^2))$-approximation using $\operatorname{poly}(1/\varepsilon)$ words of space in insertion-only streams, and $\operatorname{poly}(1/\varepsilon,\log n)$ words of space in dynamic streams. We substantially generalize and strengthen their result, provided that a highly accurate estimate of $\varepsilon$ is available. For any Boolean Max-$2$CSP and any $\eta>0$ independent of $\varepsilon$, we obtain, with high constant probability, a single-pass $(1-\eta)$-approximation using $\operatorname{poly}(1/\varepsilon,1/\eta)$ words of space in insertion-only streams, and $\operatorname{poly}(1/\varepsilon,1/\eta,\log n)$ words of space in dynamic streams, where $n$ is the number of variables. We further extend our techniques to general Boolean Max-$k$CSPs for $k\ge 3$, with slightly worse space complexity.

Federated Multi-Agent Systems (FedMAS) offer a promising paradigm for collaborative reasoning among LLM-based agents while keeping raw client data decentralized. However, reliable multi-round collaboration remains challenging. Agent selection must take into consideration evolving task requirements and the agent's historical reliability, while privacy rules must adapt to changing collaboration contexts and meet the client's personalized needs. To address these challenges, we propose PFedMA, a federated multi-agent framework for feedback driven multi-round collaboration under heterogeneous privacy requirements. Across rounds, PFedMA maintains historical capability and reliability states to guide agent selection, evaluates task responses before aggregation, and refines subsequent queries according to accumulated collaboration feedback. PFedMA enables bidirectional local to global and global to local private rule evolution across heterogeneous clients. Clients generalize locally updated privacy rules and selectively incorporate globally maintained rules, thereby facilitating cross-client privacy rule evolution while preserving personalized constraints. Extensive experiments across reasoning, privacy sensitive, and adversarial settings show that PFedMA improves task performance, reduces measured privacy leakage, and improves system robustness.

Expected-reward maximization in reinforcement learning concentrates probability mass on a few outcomes, even in tasks that require many distinct, high-quality solutions, a failure known as mode collapse. Inverse Probability Scaling (IPS) counters this in group-based policy gradients such as GRPO by dividing each reward by the outcome's probability, estimated from its frequency within the sampled group. When the outcome space is large relative to the group, outcomes rarely repeat and these estimates become uninformative. Since an outcome can be constructed along many trajectories, its probability is a sum over all of them, and computing it directly is intractable. We propose Marginal IPS (MIPS), which performs marginal evaluation by combining the sampled trajectory with a learned model of the trajectories that lead to each outcome. On Hypergrids, where the exact sampling distribution is computable, MIPS stays near the reward-distribution target as the outcome space grows relative to the rollout group, while frequency-based IPS degrades. In molecular synthesis and phylogenetic inference, MIPS sustains diverse, high-quality sampling where GRPO and frequency-based IPS collapse, showing that MIPS serves as a reliable solution to mode collapse in RL.

The problem of Private Information Retrieval (PIR) across graph-based replicated systems has seen a surge of interest in recent literature. In such systems, servers are modeled as the vertices of a graph, and each file is represented by an edge connecting the pair of servers that store it. While previous research on Graph-PIR (G-PIR) has focused mostly on perfect information-theoretic privacy, relaxing this constraint has the potential for significantly higher retrieval rates. In this work, we consider the setting of Graph-based Weak Private Information Retrieval (G-WPIR), under an $\epsilon$-privacy metric. We adapt an existing G-WPIR scheme for arbitrary graphs to analyze its performance under the $\epsilon$-privacy metric. Additionally, we establish a lower bound on the expected download cost for arbitrary graphs under the $\epsilon$-privacy metric and derive from it an explicit rate upper bound for complete graphs. Finally, we show the achievable rate-privacy trade-offs for two specific graph classes—complete graphs and complete bipartite graphs.

Digital advertising is an important means for platforms to reach potential users and drive business growth. Advertising delivery is the core link between marketing objectives and actual returns. Existing methods have achieved partial automation in individual tasks such as bidding and budget allocation, but they still rely on human operators for overall coordination, making it difficult to support large-scale operations. They also often fail to achieve optimal performance, and individual experience remains difficult to turn into reusable operational capabilities. In addition, as advertising markets change, advertising delivery systems need timely strategy updates to remain effective in dynamic markets. To automate the entire advertising operations process and improve adaptability to market changes, we propose AdCoEvo, a co-evolving multi-agent advertising delivery system. Specifically, the system connects delivery decisions across different timescales and operational granularities through role specialization. Diagnostic results guide account planning and strategy adjustment in the next period. The multi-agent system updates its model parameters based on actual delivery outcomes and continually improves delivery strategies through mutual feedback among planning, adjustment, and diagnosis, thereby achieving co-evolution. Experiments show that, relative to the rule-based operational baseline (Rule), AdCoEvo increases advertiser value by 14.02% while reducing cost per acquisition by 7.07%. Compared with a multi-agent system with fixed parameters, AdCoEvo increases advertiser value by 3.96% and reduces cost per acquisition by 1.24%, indicating that co-evolution can yield greater delivery returns.

Several recent studies propose detecting prompt injections from a language model's internal state. However, these methods do not measure a single security property. Residual-stream probes, attention scores, role projections, exposure probes, and concept readouts may capture injection presence, recognition, information routing, or behavioral influence. It is therefore unclear what differences in prompt-injection detection performance reveal about the internal signals these methods read. We study how five representation-reading methods vary under a common structured indirect prompt-injection setup on Qwen3.6-35B-A3B and Gemma-4-26B-A4B-it. We evaluate each method by how well it distinguishes injected prompts from benign controls. We then further analyze each detection method's ability to detect attacks in cases where they are successful and in cases where they are not. This helps us better assess what properties each detection method might rely on. Methods trained to detect injection presence or attack-related concepts generally separate unsuccessful injections more strongly from benign inputs, whereas attention diversion, exposure, and role-confusion-based methods separate successful injections more strongly on both models. Attention Tracker provides the strongest consistent successful injection separation, reaching 0.949 AUROC on Qwen and 0.887 on Gemma. These results suggest that signals closer to response formation may better encode information about an attack's success, and that detection performance can depend strongly on the prompt template. We share the data associated with this study at https://anonymous.4open.science/r/paper-evaluation-data-7877

We study two explicit constructions of binary Calderbank–Shor–Steane quantum low-density parity-check (QLDPC) codes with check weight six. The first uses a fixed circulant shift pattern and yields parameters $\code{6N}{2N+2}{6}$ with Tanner graph girth twelve for $N=31$ and every $N\ge37$. A graph incidence interpretation gives the ranks and girths, and a uniform weight six logical operator establishes the exact distance. The second construction uses fixed matrix polynomials with circulant and affine permutation coefficients. Here, a polynomial commutation identity guarantees orthogonality for every cyclic lift. Four realizations give $[[192,16,8]]$, $[[288,16,12]]$, $[[384,16,14]]$, and $[[576,32,20]]$, with Tanner girth eight and computationally certified distances. We give reproducible construction data, distinguish classical cycle distance from quantum distance, and identify the limitations of both families.

While Large Language Models (LLMs) show promise for Automatic Program Repair (APR), existing benchmarks suffer from data contamination, in which test samples appear in training sets, artificially inflating performance. To address this, we propose BenchChecker, a framework compatible with both open-source and closed-source models that supports more credible evaluation by detecting such contamination. Grounded in the observation that persistent patches likely remain in recent repository snapshots used for training, BenchChecker employs a forensics-based two-stage detection process using only textual model outputs together with public repository history. The first stage conducts a repository presence test by probing the model with code prefixes to infer training inclusion; the second stage performs a patch presence test to verify patch persistence via similarity alignment and test-suite verification. Evaluations using StarCoder-15B, StarCoder2-15B, and CodeGen2-16B trained on The Stack v2 demonstrate that BenchChecker achieves robust detection with AUC scores around 0.9 across six benchmarks. Furthermore, applying BenchChecker to SWE-bench Verified and Multi-SWE-bench reveals that filtering contaminated samples reduces the reported resolution rates of most LLMs by more than 20% relative to the original rates on medium-difficulty tasks.

Recommendation systems, systems that predict a user's preference for items in order to curate that user's feed, are a ubiquitous and central feature in online social networks. While they are a central piece of our modern digital infrastructure, many modern approaches are opaque black boxes, making them difficult to audit, understand, and assess the trustworthiness of. In this paper we present an exact, graph-based approach for generating explanations for decisions made by recommendation systems. We argue that in the risk-sensitive sociotechnical context of user-facing AI such as recommendation systems, explainability is a key pillar for establishing the trustworthiness of models and validating their effectiveness. We propose a technique built on LightGCN like architectures, an influential and popular class of recommendation models. We describe our approach and argue that it provides an appropriate assurance mechanism for these AI systems leading to specific advantages, including interpretable explanations and a theoretically guaranteed measure of explanation fidelity. We conclude by discussing limitations and directions for future research.

Neural instrument synthesizers can clone the timbre of a solo instrument, letting music producers make music with the sound they desire. However, an instrument of interest is often embedded in a fully produced, multi-instrument recording, beyond the reach of existing methods. Source separation before timbre cloning is an approach, but it may suffer from separation artifacts and is also likely a detour for the task. In this work, we propose Point-and-Play, a system that clones the timbre of a cued source directly from a polyphonic mixture and uses it to synthesize new pitches and velocities. To achieve this, we introduce a Cue-Gated Statistics Pooling (CGSP) module that extracts the timbre code of the cued instrument from adapted CLAP embeddings of the music mixture. This extracted timbre code is then used to condition a separately trained note synthesizer together with pitch and velocity inputs. Experiments show that our proposed system outperforms separate-then-clone baselines in terms of timbre fidelity and synthesis quality, achieving comparable performance with that cloned from a clean solo instrument.

Random reshuffling stochastic gradient descent (RR-SGD) is widely used for finite-sum optimization in machine learning. For convex $L$-smooth components, recent work establishes convergence guarantees for an averaged iterate with component stepsizes as large as order $1/L$. However, the final iterate is a natural output in practice, and existing last-iterate guarantees in this setting require stepsizes that scale inversely with the number of components $n$. We establish an expected objective-gap bound for the last iterate of standard RR-SGD on unconstrained problems with a constant component stepsize satisfying $0<\eta\lesssim 1/(L(1 + \log(nK)))$, where $K$ is the number of epochs. By retaining the sampling bias at random comparison points, we remove the explicit $1/n$ factor from the stepsize restriction. Combined with existing small-step bounds, our result matches the corresponding averaged-iterate convergence rate up to logarithmic factors.

Native multimodal pre-training trains a single model from scratch on text and multimodal data, enabling it to understand both language and visual inputs. However, how visual signals shape scaling behavior and support capabilities beyond visual understanding remains unclear. We investigate this question through controlled visual-content ablations across model sizes and training budgets. Specifically, we replace image embeddings with content-free padding token embeddings. Retaining visual content improves text prediction on image-text examples without materially changing the lowest text-only loss at a given compute budget, the optimal balance of model size and training data, or aggregate language scores. The relative multimodal loss benefit grows approximately as a power law of compute. To test benefits beyond understanding images, we evaluate spatial reasoning from text descriptions alone. Both conditions improve with model and data scaling. During training of the largest model, visual gains are most sustained in localization, navigation, and symbolic planning, which involve maintaining object configurations or composing spatial updates; gains in spatial relations, geometry, and physical commonsense are less consistent. These findings identify visual content as a scalable anchor for multimodal learning, with benefits that extend beyond visual inputs while preserving aggregate language performance.

In this paper, we develop a Koopman operator framework for modeling and control of traffic flow governed by the Lighthill–Whitham–Richards (LWR) conservation law. We show that the nonlinear dynamics induced by the Greenshields flow–density relation naturally motivate a dictionary of polynomial observables, while the controlled dynamics become bilinear in the lifted Koopman coordinates. A reduced-order Koopman model is identified from spatio-temporal traffic-density snapshots and used as the prediction model in a model predictive control (MPC) framework for ramp metering. The resulting Koopman-MPC is compared with two predictive control baselines that use, respectively, a fixed linearized LWR model and the nonlinear finite-volume LWR model. In a one-hour freeway experiment, the Koopman-based controller maintains the traffic density below the critical Greenshields value and closely reproduces the behavior of the nonlinear PDE-based MPC, with a maximum-density difference below 1%. The Koopman predictor also reduces the mean online model-evaluation cost by approximately one order of magnitude relative to the nonlinear PDE predictor. In contrast, the linearized LWR predictor is computationally cheaper but exhibits substantially larger prediction errors and yields a more conservative metering policy. These results demonstrate that the analytical structure of the nonlinear traffic conservation law can be exploited to construct an interpretable data-driven surrogate for efficient traffic prediction and control

Multimodal language models often need to reason about the physical dynamics of the world in order to answer questions appropriately. Various methods have emerged for solving physical reasoning tasks, and in particular, two leading approaches are Chain-of-Thought reasoning and physics-grounded simulation. However, relatively little is known about which particular problems can be solved by each approach. Some tasks may benefit from computationally expensive generative video simulations, whereas others may be solved equally well (or even better) with language-based reasoning. Our work addresses this by systematically comparing simulation and language reasoning across a variety of multimodal tasks involving physical reasoning, such as answering questions about fluid dynamics and elasticity. We draw a parallel with this to resource-rational theories of human cognition, which propose that people optimally trade off simulation and reasoning depending on task demands. We present a novel cognitively-inspired hybrid model which optimally trades off simulation and reasoning to improve model performance and computational efficiency. Our model is able to decide the appropriate method to solve a physical reasoning problem at test time, and finds an optimal balance between accuracy and cost. Further, our model can be trained on a smaller vision-language model that is only 4B parameters and deployed to improve performance on much bigger base models with $>$3T parameters like Gemini 3 Pro.

While LLM-based agent systems have demonstrated remarkable proficiency in reasoning and solving complex real-world tasks, they also come with substantial token consumption and inference latency, limiting their practical deployment. To enhance the efficiency of agent systems, recent efforts have focused on transferring agentic capabilities from large-scale models to small language models (SLMs) via supervised fine-tuning-based distillation. Nevertheless, as interaction trajectories grow longer, involving extended reasoning-action chains and extensive environmental feedback, distilled student agents tend to experience progress stagnation: becoming trapped in unproductive loops characterized by persistent failed strategies and difficulty making meaningful progress. Our systematic analysis shows that this phenomenon consistently occurs across diverse SLMs and task domains. To address this issue, we propose Progress-Aware Distillation (ProD), which explicitly detects and penalizes stagnation in student-generated trajectories. ProD encodes the stagnation gap between student and teacher into an adaptive margin for dynamically penalizing possible stagnation. By iteratively applying this procedure, ProD enables student agents to progressively approximate teacher behavior. Extensive experiments involving 8 SLMs, ranging from 0.6B to 8B parameters, on both in-domain and out-of-domain benchmarks, demonstrate that ProD substantially mitigates progress stagnation while enhancing both the success rates and task-completion efficiency of distilled SLM agents.

Vision-language-action (VLA) policies fine-tuned on LIBERO are trained and evaluated on the same ten instruction strings, and rewording the instruction changes base OpenVLA's success by -38.0 points. Test-time prompt ensembling, which improves language models, does not recover this loss in any setting we ran. Paraphrase consensus (the medoid action over K paraphrases) costs base OpenVLA 16.8 points when the training instruction is a member. Without it, the medoid of the paraphrases scores -32.4 points relative to the training instruction, and confidence-based selection scores -28.0. The same medoid over samples of one paraphrase improves on greedy decoding but stays -27.6 points below the training instruction, consistent with aggregation reducing variance without removing the bias the paraphrases share. A wrong-instruction probe measures how much of a policy's success depends on the instruction. Across three policies and four suites, paraphrases cost a policy only where this share is high, and the stronger policies lose a much smaller fraction of it than base OpenVLA. LoRA post-training on paraphrased instructions reduces the one-paraphrase effect to +2.0 points, against -32.0 for a matched control, reproduces across two seeds and on OpenVLA-OFT, transfers to withheld phrasings, and leaves the policy's reliance on the instruction unchanged.

Unsupervised feature selection (UFS) is of great significance in image processing and pattern recognition. Nevertheless, existing model-based methods often rely on manually tuned hyperparameters, while deep learning-based methods generally require massive training data. To address these issues, this paper proposes an efficient manifold deep unfolding feature selection network (MDUFS-Net). Technically, we first incorporate the $\ell_{2,0}$-norm into the classic principal component analysis (PCA) framework, then unfold the alternating direction method of multipliers (ADMM)-based iterative algorithm into an end-to-end trainable network, thereby enabling automatic parameter learning. Moreover, a manifold optimization strategy is developed to improve the numerical robustness. On this basis, we further extend MDUFS-Net into an enhanced variant called MDUFS-Net+ by introducing the $\ell_{0}$-norm for more discriminative feature representations. Extensive experiments validate that our proposed MDUFS-Net and MDUFS-Net+ outperform state-of-the-art methods in the UFS task, with promising performance in clustering. Our code is available at https://github.com/xianchaoxiu/MDUFS.

We develop and analyze a three-dimensional pure source transfer domain decomposition method (PSTDDM) for the Helmholtz equation in free space. Extending the two-dimensional PSTDDM proposed by [Du and Wu, J. Sci. Comput., 83(2020),83:68], the proposed method transfers sources successively along the $x_1$, $x_2$, and $x_3$ directions, yielding layer-wise, bar-wise, and block-wise PSTDDMs. The main challenge in the convergence analysis is deriving wavenumber-explicit truncation-error estimates. To address this challenge, we establish inf–sup conditions for the uniaxial perfectly matched layer (UPML) problem under relaxed restrictions on the PML parameters and reformulate the source-transfer operators. We then introduce a positive-definite reconstruction step that produces a globally conforming approximation. In addition, a new quantitative lower bound is established for the imaginary part of the complex PML distance, more easily yielding exponential decay estimates in three dimensions. These lead to wavenumber-explicit convergence estimates for all three PSTDDMs. Numerical experiments demonstrate the effectiveness of the proposed methods both as direct solvers and as preconditioners for GMRES.

Static benchmarks for large language model (LLM) agents face saturation and data contamination risks, motivating continual updates to evaluation tasks. However, aggregate scores and success rates offer limited insight into specific problem-solving bottlenecks and provide limited guidance for subsequent task design. We propose EvoProbe, an evaluation-driven benchmark evolution framework that integrates task generation, diagnostic evaluation, and feedback-driven regeneration into a closed loop. It combines verifiable subtask outcome scoring with evidence-supported multi-judge trajectory evaluation and distills phase-level bottlenecks, success patterns, and failure patterns into diagnostic feedback for generating new executable tasks. Across two rounds of cybersecurity experiments, evaluations of eight models on 66 tasks reveal phase-level performance differences and common failure patterns beyond aggregate scores. In a cross-round comparison with the Examiner model set and solver configuration held fixed, all eight models achieve lower Outcome scores on the updated tasks, with relative decreases of 11.8% in mean Outcome and 40.0% in mean Pass\textasciicircum{}3. These results support using fine-grained diagnostics to guide iterative task updates and probe agents' current capabilities and limitations.

Comparing financial disclosures across languages requires identifying shared accounting concepts and recovering all numerical facts that express them. Existing benchmarks evaluate aspects of these capabilities, but provide limited insight into whether both succeed on the same multilingual evidence. In particular, it remains unclear whether language models can recognize these concepts across languages and exhaustively recover their corresponding numerical evidence from the same contexts. We introduce MULTIFINGROUND, a benchmark designed to investigate this recognition-grounding gap in multilingual financial reports. Constructed from FY2025 disclosures of 90 issuers in English, French, and Greek, MULTIFINGROUND aligns independently authored reports through a shared accounting taxonomy and provides traceable fact-level annotations. The benchmark evaluates financial fact identification, concept recognition, exhaustive fact selection, and concept discrimination without candidate labels. Across 15 evaluated models, eight API and open-weight reference models achieve 97.3-99.4% concept recognition accuracy, yet miss at least one target fact in 26.6-41.9% of correctly recognized groups. Further analyses reveal substantial variation in concept discrimination and limited gains from the evaluated finance-specialized models. Our findings demonstrate that strong concept recognition can coexist with incomplete evidence recovery, highlighting the importance of evaluating semantic recognition and evidence completeness together.

This paper presents $S^3$-Nav, a structured spatial-semantic decision-making framework for language-guided robot navigation under partial observability. Rather than assuming a complete environment representation in advance, $S^3$-Nav incrementally constructs persistent spatial-semantic state while exposing only a capacity-controlled, task-conditioned view to the high-level policy. This separation decouples what the robot must retain from what the policy needs to inspect, allowing decision complexity to be controlled without discarding grounded evidence required for subsequent reasoning and validation. High-level navigation outputs remain provisional and are independently grounded and validated against persistent semantic evidence and current geometric observations before physical execution or task termination. Evaluations across multiple environments and object- and relation-goal tasks demonstrate that separating persistent state, decision-facing evidence, and execution authority improves closed-loop navigation reliability while reducing policy reasoning cost.

We propose a potential projection (PP) method for the nonlinear Schrödinger equation (NLSE) with low-regularity and singular potentials by replacing the potential with its Fourier projection (truncated Fourier series). This gives rise to an NLSE with projected potential (NLSEwPP). We rigorously establish its well-posedness under low-regularity potentials, and its convergence to the original NLSE. The NLSEwPP is then directly discretized by the standard time-splitting Fourier pseudospectral (TS-FP) method with the mesh size chosen according to the potential projection parameter, resulting in the fully discrete TS-PP-FP method. Numerically, TS-PP-FP consistently achieves optimal temporal and spatial convergence rates with respect to the regularity of the exact solution under low-regularity and even singular potentials. Compared with the time-splitting Fourier spectral method, TS-PP-FP attains the same convergence rates at significantly lower computational cost and preserves the mass conservation and time symmetry (that are lost in the Fourier spectral discretization of the time-splitting method); compared with the time-splitting FP method, it is substantially more accurate for low-regularity potentials with the same computational cost.