Papers

We introduce CT2Yarn, a human-in-the-loop framework for recovering a single continuous yarn path from micro-computed tomography (micro-CT) scans of real crochet objects. Crochet is a craft that creates complex three-dimensional shapes by interlocking loops formed from a single yarn. Recovering the underlying yarn path from external observations is challenging because of severe self-occlusion. While micro-CT reveals the full internal structure of a crochet object, the volumetric scan alone does not explicitly encode how the yarn traverses the object. This challenge stems from the hierarchical structure of yarn: a yarn consists of multiple twisted plies, and each ply itself consists of twisted fibers. Consequently, local fiber orientations observed in micro-CT scans are not aligned with the overall yarn direction. To recover yarn-level orientations from ply-level fiber orientations, we first estimate local fiber directions using Gabor filtering and convert the volume into an oriented point cloud. We then introduce an anisotropic mean-shift procedure that aggregates local fiber orientations within a neighborhood of the yarn radius into yarn-level orientation estimates. Combined with an automatic topology skeletonization strategy, our method extracts yarn-path fragments. Subsequent fragment linking, junction cleaning, and loop detection merge these trees into a small number of long curves. Where the automatic reconstruction remains ambiguous, a sketch-based user interface enables users to interactively complete the single continuous yarn path. The recovered yarn path enables downstream applications including physically based simulation, ply-level rendering, and stitch-pattern extraction.

Synthesizing photorealistic driving videos along specified trajectories is essential for scalable closed-loop simulation. Reconstruction-based methods leverage neural rendering to synthesize geometrically consistent views, but often exhibit diverse artifacts and missing content when the viewpoint deviates from the training trajectory. In contrast, generative models can synthesize realistic views along arbitrary trajectories from vehicle sensor data, yet often struggle to maintain temporal and geometric consistency across frames. To combine the strengths of both, we propose PRG-Fusion, a framework for driving view synthesis that uses reconstruction evidence to orchestrate generative priors across regions. Specifically, we extract region-wise degradation evidence from reconstructed driving scenes and convert it into Preserve, Repair, and Generate (PRG) labels. At inference, these labels serve as a unified routing policy for region-aware spatiotemporal synthesis, orchestrating 3DGS appearance preservation, LiDAR-guided structural correction, and video-prior-driven content completion across Preserve, Repair, and Generate regions, respectively. We then follow a two-stage training paradigm, first establish geometric control from sparse LiDAR projections and subsequently learning appearance control from dense 3DGS renderings. Extensive experiments on Waymo demonstrate that PRG-Fusion achieves state-of-the-art overall performance in novel trajectory video synthesis, with superior visual quality and geometric fidelity while maintaining competitive view consistency under large trajectory shifts.

Flow matching, together with classifier-free guidance (CFG), is widely used in generative modeling, yet much of the theoretical understanding remains distribution-wise. Since practical sampling follows individual trajectories, distribution-level guarantees alone do not fully capture how trajectories interact with the data geometry or how guidance reshapes it. To overcome this limitation, we establish a unified stagewise geometric theory of attraction and absorption for both continuous dynamics and explicit Euler discretization. Specifically, with $t\in[0,1]$ running from noise to data, we show that unconditional flow trajectories are successively attracted toward a neighborhood of the global mean, the data convex hull, and a neighborhood of a possibly nonconvex local cluster. Across these stages, the corresponding distance satisfies a common contraction estimate, yielding an ${O}(1-t)$ decay of the distance in the final stage. For CFG, the same structure persists with an extrapolated mean, an inflated conditional convex hull, and, near the target cluster, the restored local geometry of conditional flow matching. We further show that a general time schedule $a(t)$ replaces the $O(1-t)$ decay by $O(1-a(t))$. Together, these results provide a unified particle-level geometric account of flow matching and CFG across continuous and discrete sampling.

Medium-range precipitation forecasts are impaired by persistent systematic biases, lead-time-dependent error accumulation, and coarse spatial resolution, restricting their reliability for flood-drought risk assessment. Existing AI correction techniques lack dedicated modeling for multi-day dynamic bias evolution and proper meteorological constraints, often generating over-smoothed rainfall structures, and cannot meet operational deployment demands. This work introduces PCSDiff, a cascaded task-decoupled diffusion framework targeting 10-day precipitation bias correction and downscaling. To jointly counteract temporal error drifts and reconstruct physically plausible local precipitation details, PCSDiff integrates the Precipitation Intensity-aware Multi-branch Decoder (PIMD) module for dynamic multi-day error mitigation using synoptic-temporal features, followed by a two-phase conditional diffusion super-resolution module to restore fine-scale precipitation patterns. Evaluated against CMA-CRA observations over China after global-data training, PCSDiff cuts RMSE by 16.1% and lifts ACC by 13.9% relative to raw ECMWF forecasts at 3-10-day lead times, and consistently outperforms mainstream deep-learning baselines on both general and extreme-precipitation metrics. Benefiting from a streaming inference pipeline, our method achieves low-latency rolling forecasting for practical meteorological operations.

Causal effect estimation asks how an outcome would change under an intervention, and medicine, economics, and public policy all treat it as a foundational task. Prior-data fitted networks (PFNs) amortize the task: a model trained on large numbers of programmatically generated synthetic causal tasks reads a new problem's observational data into context and returns an interventional-effect estimate in a single forward pass. The capability of such models is largely determined by the synthetic training prior, which is currently designed by hand, a bottleneck acknowledged by both Do-PFN and CausalPFN. Large language models (LLMs) can now “draw” plausible causal graphs for a given domain, suggesting that LLM-distilled graphs could serve as prior material. Whether injecting such graphs helps at all, where any gain comes from, and when injection helps. Practice has so far relied on manual trial and error. We propose a closed-loop prior selection framework that casts prior injection as a budget-constrained optimization over a candidate prior pool. Candidates undergo cheap post-training and are scored by a composite metric dominated by real-domain generalization; the winner then receives full training and paired statistical validation. On a 7.34M-parameter Do-PFN, the framework's winner attains a formally significant $2.75\times$ gain on the primary evaluation domain, and its error falls below that of the uninjected official base. Generalization on an adjacent monitoring domain improves significantly, and no monitored capability degrades. Mechanism experiments show that the gain depends on the semantic content of the distilled graph rather than its structural diversity alone does not produce it (directional evidence). With this framework and this regularity in hand, the use of LLM causal priors stops being manual trial and error and becomes an empirically verifiable selection problem.

Large language model (LLM) inference increasingly spans models that differ in size, capability, price, and provider. This shift creates two costs for developers. One is the integration cost of choosing among and switching between many models. The other is the inference cost of rebuilding a KV cache when it is unavailable or incompatible with the selected model. We define and analyze the Unified AI Gateway as a system setting for an edge-deployed AI traffic hub. It coordinates model routing, KV cache management, and compute placement across end devices, edge resources, and cloud model services. At request time, the gateway jointly selects a target model, an execution site, and a KV cache action under task-quality, latency, cost, and resource constraints. In parallel, background cache-management actions optimize KV cache placement, replication, retrieval, and lifecycle decisions for subsequent requests. We synthesize existing evidence on KV cache reuse, compression, cross-model mapping, distributed storage, and transfer, and discuss the remaining challenges of integrating these capabilities into one system. Across eight typical workload profiles, our workload-level analytical simulation reports TTFT speedups of 1.25$\times$–13.28$\times$ and input-cost benefits of 1.20$\times$–6.16$\times$.

Post-hoc safety training (RLHF, DPO) is the dominant way to align large language models, yet jailbreaks (Zou et al., 2023), fine-tuning attacks (Qi et al., 2024), and activation-space edits (Arditi et al., 2024) keep recovering the behaviors it was meant to remove. We give this fragility one geometric explanation and follow it into pretraining. We measure the safety update $\Delta = W_{safe} - W_{base}$ against the curvature of the model's capabilities (the empirical Fisher of a capability loss). Across five model families, post-hoc safety lands in a suppression regime: $\Delta$ is nearly orthogonal to the capability directions, and its small in-subspace part concentrates on a few high-curvature ones. The update is thin but sharp, a refusal gate laid over intact capabilities rather than erasure of them. A kernel-immobility lemma explains why such an update can only mask a capability, not remove it, and a little benign fine-tuning switches the gate off: 100 benign examples cut the AdvBench refusal of Qwen-2.5-7B-Instruct and Llama-3-8B-Instruct by 35 to 38 pp. Following the account into pretraining, a pretraining-checkpoint sweep of OLMo-2-1B (Team OLMo et al., 2024) shows the features that refusal attaches to emerging in a sharp transition between 1B and 63B pretraining tokens. We then use the account constructively: models trained from scratch with safety co-training spread continuously across pretraining reach 87 to 98% AdvBench refusal that the same attack erodes by only 2 to 14 pp at every scale from 410M to 6.9B, at a small cost on short-answer capability probes; a windowed schedule of equal total safety weight installs no refusal. Persistence, not timing, decides whether refusal survives. It does not keep output safer: by a harmful-output classifier, the attack raises harmful completions by 8 to 12 pp after post-hoc safety and by 11 to 27 pp after co-training.

Invasive coronary angiography (CAG) is the gold standard for diagnosing coronary artery disease, but interpretation varies substantially among observers. Existing AI systems can improve consistency but lack auditable decision processes and are limited in comprehensive open-ended assessment, undermining clinician trust and clinical adoption readiness. We developed CARDEA, a unified large vision-language model that serves as the inference core of a CAG pipeline. It was trained solely on public datasets and closed-ended tasks in three stages: visual feature alignment, a self-distilled Chain-of-Box (CoB) cold start, and reinforcement learning with verifiable rewards (RLVR) with a CoB reward encouraging bounding-box use in the reasoning trace. We assessed its two study-level diagnoses, dominance classification and complexity assessment, against a dedicated classifier and two interventional cardiologists. Report generation was excluded from training and evaluated zero-shot across stages on an external cohort using vessel-severity macro-$F_1$. CARDEA trailed the classifier on in-distribution dominance but drew level under domain shift (accuracy, 0.91 [95% confidence interval (CI), 0.86 to 0.95]) and was comparable to the cardiologists on complexity assessment (accuracy, 0.90 [CI, 0.82 to 0.97]). Only RLVR improved zero-shot report generation, raising its vessel-severity macro-$F_1$ (0.686 [CI, 0.664 to 0.707]) above the untuned base model (0.513) and over twice the always-normal floor (0.312). CARDEA runs an end-to-end CAG pipeline from raw multi-view videos through keyframe selection to study-level diagnosis while exposing auditable spatial evidence behind its conclusions. RLVR on verifiable closed-ended tasks surfaced open-ended reporting ability that supervised imitation did not. Clinical use requires prospective validation against expert cardiologists.

Distributed Dexterous Manipulation (DDM) is a novel paradigm that presents significant control challenges due to high action-space redundancy, inter-robot cooperation, and dynamic object-robot interactions. This paper introduces a framework based on spatially conditioned Multi-Agent Transformers (MATs) to efficiently learn robust control policies for a DDM system grounded in an array of 64 soft delta robots arranged in an 8x8 grid. Our three core contributions are: (i) an MAT with adaptive layer norm for compute efficiency, (ii) spatial contrastive embeddings to ground transformer embeddings in the spatial configuration of the robots, and (iii) an MAT-based behavior cloning method fine-tuned using Soft Actor Critic. We also propose an action selection formulation to analyze the trade-off between task performance and the number of robots utilized. Our experiments show that MATs iteratively refine their actions through the stacked attention blocks. This further informs the benefit of spatial conditioning in transformers to learn DDM policies. We demonstrate long-horizon planar manipulation tasks with objects of various geometries in simulation and real-world. Finally, we show how action selection mitigates robot maintenance by reducing wear and tear due to inter-robot collisions while maintaining the ability to manipulate objects along various trajectories in the real-world, achieving an average error of ~1.5 cm, while using ~65% fewer robots.

Inter-frame affine misalignment caused by platform jitter and attitude adjustments poses a fundamental challenge for multi-frame analysis of point targets in optical surveillance. Conventional registration methods rely on spatial intensity correlations or distinctive image features, both of which are largely absent in low-signal-to-noise-ratio point target imagery. We introduce the Radon Point Spread Function (RPSF) to characterize point targets in the Radon-transformed domain, and derive a closed-form framework that jointly estimates inter-frame translation and rotation from as few as four scalar RPSF samples per frame pair. The method requires no iterative optimization, feature extraction or interpolation, which is suitable for resource-constrained onboard processing. Simulation results confirm sub-pixel translation accuracy and a mean rotation error of 0.2556{\deg} at 1{\deg} Radon angular resolution. Validation on five real space debris datasets including both ground-based and in-orbit observations yields a mean calibration error below 0.5 pixels, substantially exceeding the precision required for reliable multi-frame processing.

Emotion analysis is a fundamental task in computer vision, but its practical deployment remains constrained by the privacy risks inherent to conventional RGB cameras. Bio-inspired event cameras present a promising hardware-level solution because they capture asynchronous brightness changes, thereby reducing exposure of facial identity details while leveraging high dynamic range for robust perception under challenging illumination conditions. Despite these advantages, existing event-based methods struggle in complex real-world settings due to limited dataset scales, simple acquisition conditions, and reliance on single-modality visual cues. To address these, we establish a challenging tri-modal benchmark with event, audio, and text modalities and propose the Information-Guided Gated Fusion (IGF) framework, which first pre-trains an event encoder on the FAU subset of Emo-DVS to capture fine-grained facial dynamics, then employs adaptive modality gating to suppress modality-specific noise, and finally leverages mutual information maximization to align robust cross-modal representations. To alleviate data scarcity, we introduce Emo-DVS, the first large-scale event-based emotion analysis dataset, which couples dynamic illumination with the Facial Action Unit (FAU) subset and emotion subset. Extensive experiments demonstrate that IGF achieves state-of-the-art performance.

For a stream of nonzero vectors $x_1,\ldots,x_n\in\mathbb{R}^d$, let $u_i=x_i/\|x_i\|_2$. We define the cosine density of the $i$-th stream element by $D_i:=\sum_{j\in[n]}\langle u_i,u_j\rangle$ and study the density moments $M_p:=\sum_{i\in[n]}D_i^p$ in both the signed- and nonnegative-cosine regimes. These quantities are similarity-aware analogues of classical frequency moments: replacing cosine similarity by equality (that is, $D_i = \sum_{j\in[n]} \mathbf{1}\{u_j = u_i\}$) gives $M_p=F_{p+1}$ and, in particular, $M_{-1}=F_0$, the number of distinct elements. We give one-pass streaming algorithms and lower bounds whose dependence on $d$ is tight up to logarithmic factors for $p=-1$, $p=1$, and even $p \ge 2$ (the latter under signed cosine), and whose exponents differ by less than $1/2$ for odd $p \ge 3$ under nonnegative cosine. Our results thus extend several fundamental statistics from the classical data stream literature to cosine similarity, a widely used measure for comparing vector embeddings in modern AI systems. The main challenge in proving a space lower bound for nonnegative cosine is to eliminate unwanted contributions without relying on pairs of opposite vectors. We address this through a construction that we call equal-sum moment isolation: two insertion-only prefixes have the same cardinality and vector sum, and a finite-difference comparison cancels the baseline shared by all their vectors while isolating the desired higher-order signal. This proof framework may be useful for other insertion-only streaming lower bounds, where direct cancellation is not possible.

Automated parking requires generating complete and executable trajectories in highly constrained spaces with low tolerance for goal pose error. Existing end-to-end parking methods struggle to jointly achieve inference efficiency, trajectory quality, and precise endpoint alignment, while conventional imitation objectives provide limited supervision on structured deviations from expert maneuver geometry. We propose DriftParking, a one-step trajectory generation framework that reconstructs the drifting-field paradigm for high-precision conditional trajectory generation. Specifically, we replace distribution-level attraction with conditional one-to-one attraction toward the paired expert trajectory, introduce expert-centered constructive repulsion, and adaptively attenuate repulsion near convergence. We further formulate trajectory generation in an endpoint-residual space by decomposing each trajectory into a start-to-goal baseline and a learnable residual, turning endpoint alignment into a representation-level structural constraint on the supervision target while providing a structured space for repulsive supervision. DriftParking achieves state-of-the-art performance across all evaluation metrics. Closed-loop on-vehicle experiments across diverse parking scenarios further show a 97% parking success rate, demonstrating strong zero-shot generalization.

Quantization makes vision language models (VLMs) cheaper to store and run by using fewer bits to represent their weights. While unchanged answers on visual question answering (VQA) after compression are an expected behavior, they can still hide changes in the underlying scores (log probabilities). For example, a model may still answer yes after compression, even as the score gap between yes and no shrinks. We introduce BEFORE THE FLIP to measure these hidden changes. Our method compares the score change caused by compression with the change caused by replacing the image's internal representations, or image tokens, with one fixed average token. We then increase the precision of one weight group at a time to identify where extra bits help, and test whether choosing different groups for each question offers benefits beyond shuffled controls. Among 8,277 LLaVA questions where image token replacement measurably affects the scores, 4-bit compression shifts the yes or no score gap farther toward the replacement output than 8-bit compression. Qwen shows the same pattern, but with a smaller difference. Yet only 265 of 9,000 LLaVA answers change at 4 bits. In a separate study of 1,024 calibration questions, choosing weight groups separately for each question does not outperform both shuffled controls at any tested storage budget. These findings show that compression can alter the scores behind unchanged answers, but do not establish a reliable benefit from adjusting precision for each question.

We study constrained online convex optimization with adversarial constraints and conditionally unbiased, finite-variance observations of constraint values and gradients. Under common feasibility, our \LEDGER\ algorithm attains $O(\sqrt T)$ expected regret and $O(\sqrt{T\log(eT)})$ expected budget violation, the largest cumulative overspend over any window. It uses a reflected exponential potential, clipped signed observations, and predictable adaptive regularization, with one feedback triple and one projection per round. Neither a Slater condition, independence between feedback channels, nor an absolute constraint-value bound is needed. A Gaussian testing lower bound proves that the budget rate has optimal horizon dependence under square-root regret at fixed positive noise, including the logarithm. The same obstruction holds for terminal violation, so the logarithm is not a cost of maximizing over windows; an $O(\sqrt T)$ budget bound instead forces linear regret. In contrast, fixed positive Gaussian value noise yields a joint regret–hard-violation lower bound of $\Omega(\min\{\sigma,1\}T/\log^2 T)$, even with exact gradients in one dimension. The hard-violation construction matches arbitrarily many moments while preserving a feasible-endpoint gap and constant endpoint probabilities. Together, the bounds separate uncertainty about hard feasibility from learnable signed budgets. Deterministic restarts remove the horizon input without changing either upper rate.

Large language models (LLMs) have demonstrated strong capabilities across diverse domains, showing considerable potential in medicine. However, their application in medical settings remains limited by the scarcity of visual question answering (VQA) datasets that capture clinical reasoning and explicit image-text alignment. Here, we leverage de-identified medical images and expert commentaries shared through clinician-oriented online resources. By combining an advanced LLM with clinician-in-the-loop verification, we established a rigorous pipeline to construct ThoughtMed-1M, a long-form medical VQA dataset containing over one million VQA pairs and designed to capture structured clinical reasoning and medical image-text alignment. To demonstrate its utility, we developed a FOundational LLM Trained on ThoughtMed-1M (FOLTMed). FOLTMed achieved state-of-the-art performance across 42 medical VQA benchmark datasets, with a macro accuracy of 85.4 percent. It also generated more clinically coherent responses on the ThoughtMed-1M test set, outperforming state-of-the-art models by 3 to 5 percent across factuality and similarity metrics, highlighting a scalable paradigm for advancing research on clinically grounded multimodal LLMs.

Tabular foundation models (TFMs) are commonly pretrained on large collections of procedurally generated synthetic tasks, yet it remains unclear how well these synthetic pretraining priors support the downstream tasks on which the models are evaluated. We study this question from a distribution-level attribution perspective. We recover or reconstruct the synthetic data generators of four TFMs and compare their generated tasks with datasets from two widely used tabular benchmarks. Each dataset is represented by a common set of structural descriptors capturing schema, feature distributions, dependence structure, response properties, and feature–response relationships. In this space, we measure how broadly and repeatedly each synthetic prior reaches benchmark tasks using structural coverage and normalized density, and examine whether stronger local support is associated with better predictive performance. We find substantial differences across synthetic pretraining priors: some generators provide consistently broader and denser support for benchmark tasks than others. Moreover, stronger synthetic-to-benchmark support is generally associated with better relative model performance. These results suggest that structural coverage provides a useful diagnostic for characterizing synthetic pretraining priors and relating their data-generating assumptions to downstream model behavior.

We develop a new low-accuracy sampler, called smoothed Picard Hamiltonian Monte Carlo, which combines Gaussian smoothing, Picard iteration, and higher-order discretization. For a log-concave target $\pi \propto \exp(-V)$ in dimension $d$ satisfying $0 \prec \alpha I \preceq \nabla^2 V \preceq \beta I$, with condition number $\kappa := \beta/\alpha$, smoothed Picard HMC returns a sample with $\sqrt \alpha\,W_2(\cdot,\pi) \le \varepsilon$ using $\widetilde O(\kappa^2 + \kappa^{7/6} d^{1/6}/\varepsilon^{1/3})$ gradient queries. We also prove stronger $W_q$ bounds, and then develop an algorithmic framework, the recursive warm start generator, to upgrade these $W_q$ bounds to stronger divergence guarantees. This produces a warm start for the proximal bouncy particle sampler, introduced in a companion work, leading to a high-accuracy log-concave sampler with complexity $\widetilde O((\kappa^{7/6} d^{1/6} + \kappa^{1/2} d^{1/4})\mathrm{polylog}(1/\varepsilon))$.

We study the problem of sampling from $\mu(\mathrm{d}x)\propto e^{-V(x)}\,\mathrm{d}x$ on $\mathbb{R}^d$, where $V$ is $\alpha$-strongly convex and $\beta$-smooth, and write $\kappa:=\beta/\alpha$. We design and analyze the Proximal Bouncy Particle Sampler (Proximal BPS), a new sampler that combines ideas from the proximal sampler and the bouncy particle sampler. From a warm start initialization with $ O(1) $ Rényi divergence w.r.t. $\mu$, Proximal BPS returns a sample whose law is $\varepsilon$-close to $\mu$ in total variation distance using $\widetilde O(\sqrt\kappa\,d^{1/4} \,\mathrm{polylog}(1/\varepsilon))$ gradient queries in expectation.

Data assimilation by nudging (also called CDA) yields exponentially decaying errors and an infinite predictability horizon if the method parameter is large enough and the observations are frequent enough in time and dense enough in space. We consider the complementary case of moderate parameters and sparse and infrequent observations. We prove that assimilation with any data and any(positive) parameter strictly decreases errors and strictly increases the (now finite) predictability horizon.

Byte Pair Encoding (BPE) constructs vocabularies through greedy pair merging, but the resulting merge order does not necessarily allocate a fixed model-visible vocabulary optimally for compression. We propose Dynamic-Programming-Guided Hierarchical BPE (DH-BPE), a vocabulary-construction method that combines token exposure under exact minimum-token segmentation with the hierarchical dependencies induced by BPE training. Starting from a modestly overshot BPE candidate vocabulary, DH-BPE uses dynamic programming to measure candidate utility and applies exposure-guided, dependency-aware pruning to select a fixed-size model-visible vocabulary. We compare DH-BPE against Standard BPE and recent vocabulary-optimization baselines, including Pruned BPE, MinGram, and MinGram-PP, in primary evaluations at 12K and 16K target vocabulary sizes, with an additional 18K evaluation against MinGram only. Across the primary 12K and 16K comparisons, DH-BPE consistently improves aggregate compression over Standard BPE, Pruned BPE, and MinGram under a shared exact minimum-token DP encoder. MinGram-PP achieves stronger aggregate compression in the primary comparisons, but DH-BPE outperforms it at overshoot factors f = 2.0 and f = 3.0 in cross-corpus evaluation; at 12K, MinGram-PP reverses this ordering only with the substantially larger candidate pools at f = 4.0 and f = 5.0. Qualitative analysis further shows that DH-BPE balances later, more complete BPE merges with reusable subword components, providing a practical approach to improving vocabulary allocation under a fixed model-visible vocabulary budget.

Distributed learning control for multirobot systems (MRS) offers significant flexibility in presence of uncertainties but lacks provable performance guarantees. A promising direction involves integrating reinforcement learning (RL) into distributed model predictive control (DMPC), leveraging the strengths of RL in nonlinear policy design and the receding-horizon replanning capabilities of DMPC. However, ensuring secure control within such a learning framework under malicious cyber attacks, particularly stealthy ones, remains a critical challenge, because the distributed policies generation depends on information exchange among neighbors, where compromised agents can rapidly influence the behavior of others through the communication network. This article proposes a distributed secure learning control (DSLC) framework for large-scale MRS under malicious, stealthy actuator attacks. Our framework offers two key features: (i) a unified approach that enables secure learning control across various coordination scenarios and (ii) a game-theoretic distributed learning-based predictive control strategy that learns how to balance the attacker and defender through a differential-game based DMPC framework. Specifically, DSLC employs a distributed attacker-actor-critic architecture to learn the optimal defense and attack policies online within each prediction interval. Unlike numerical optimization-based controllers that calculate open-loop control sequences, our method simultaneously generates adversarial attack policies and corresponding defense policies in analytical closed-loop form. The defense policies could be directly generalized to MRS with varying scales and diverse actuator attack probabilities. The effectiveness and scalability of DSLC are validated through comprehensive simulations and real-world experiments in multiple wheeled robots via various control tasks.

Direct preference optimization DPO is a promising offline approach for aligning large language models (LLMs) due to its simplicity, computational efficiency, and implicit modeling of human preferences. Interestingly, iterative extensions of DPO have achieved stronger performance on academic benchmarks, raising two key questions: (i) Why do iterative methods generally outperform offline ones? (ii) Can their advantages be incorporated into offline alignment? To answer the first question, our controlled experiments reveal that the explicit preference model, additionally introduced in the iterative procedure, is a key factor behind its superiority over offline methods. This insight leads us to answer the second question affirmatively and propose Distilled Preference Probability Policy Optimization (DP3O), an effective and efficient offline alignment algorithm. DP3O first learns an explicit preference model using a helper class of LLMs and then distills its knowledge into policy optimization. Theoretically, we show that explicit preference modeling admits better estimation error control than implicit formulations, and that DP3O achieves a tighter generalization bound than hard-label DPO through variance reduction. Empirically, we evaluate DP3O on a wide range of chat-based and downstream tasks and show that it outperforms state-of-the-art offline methods, achieves performance comparable to iterative DPO, and reduces training time by about $42\%$, demonstrating both its effectiveness and efficiency.

This paper presents an enhanced formulation of the Fragile Points Method (FPM), a truly meshless approach for efficiently modeling implicit interfaces in two-dimensional differential equations involving non-homogeneous materials. The proposed framework eliminates the need for specialized numerical integration techniques and provides a systematic mathematical foundation for solving interface problems. Discontinuities in both primary and secondary variables across interfaces are naturally handled through the inherently discontinuous shape functions of FPM. Unlike conventional Galerkin methods, FPM employs simple, local, point-based polynomial trial and test functions constructed via a generalized finite difference approach. These discontinuous functions bypass the continuity requirements of standard Galerkin frameworks. To address the resulting inconsistency due to discontinuities, we incorporate numerical flux corrections inspired by the discontinuous Galerkin method. The proposed method is validated through several benchmark problems, demonstrating its efficiency and robustness.

Predictive coding characterizes adaptive behavior as a dynamic balance between internally generated predictions and external sensory evidence, yet how an embodied cognitive system can regulate this balance online during ongoing interaction remains poorly understood. This study proposes a predictive-coding-based mechanism for regulating internally generated and externally coupled processing during physical human–robot interaction. The framework employs a predictive-coding-inspired variational recurrent neural network (PV-RNN), in which a meta-prior controls the degree to which posterior inference is constrained by learned prior dynamics. We extend this architecture with an online mechanism that uses reconstruction error accumulated over recent interaction history to select between predefined meta-prior regimes. The mechanism was evaluated across three physical human–robot interaction tasks involving fixed structured, changing structured, and less-constrained interaction. Across all tasks, lower meta-prior values produced the expected increase in posterior–prior divergence and reduction in reconstruction error. More importantly, reconstruction-history-driven regime selection was also associated with reduced prospective prediction error and robot-side physical interaction conflict, demonstrating consequences beyond the retrospective reconstruction objective itself. Task~3 further showed that recent sensory observations can be successfully accommodated while subsequent human motion still departs from the model's prior-generated future trajectory. Overall, these findings show that accumulated reconstruction mismatch can provide an endogenous signal for regulating how strongly subsequent inference relies on learned internal dynamics relative to ongoing sensory input during embodied interaction.

Automated experimentation is most effective when the observables, available actions, and objective are defined before the experiment starts, as is the case for Bayesian optimization. However, in many exploratory experiments, the variables that describe the sample must be extracted from the data, new operations emerge during the experiments, and the instrument budget is too small to learn the problem by trials. Here we introduce the Scanning Probe Agentic Research Cycle (SPARC) framework, in which a coding agent and a human operator share one microscope, one notebook, and two persistent memory files. FINDINGS.md stores graded conclusions about the experiment, whereas PITFALLS.md records learned failure modes of analysis and instrument. We apply SPARC to reconfigure the in-plane superdomain direction of a (111)-oriented PbZr0.2Ti0.8O3 film. In an operator-supervised campaign, the agent reanalyzed earlier manual measurements and developed an oriented lattice of stationary bias pulses with alternating polarity to reconfigure the superdomain direction. In a subsequent agent-controlled campaign, PITFALLS.md entries were compiled into checks that validate a design before any write. The experiments showed that spatial polarity alternation, instead of the exact matching between the lattice and lamellar periods, determines directional selection. Combining a raster scan with a masked pulse lattice printed the letters UTK into the superdomain orientation. The campaign also identified practical requirements for agentic experimentation where physical verification of instrument execution, the conditions under which stored findings remain valid, validation of new observables on instrument data, and robust control protocols.

Diffusion policies offer a powerful and expressive parameterization for continuous control. Yet, their integration with reinforcement learning remains conceptually and algorithmically challenging. In this work, we address this gap by introducing a noisy-space action-value (Q-)function that assigns values to diffusion latents through the distribution of executed actions induced by the denoising process. We show that this construction admits a precise semantic interpretation and derive a noisy-space policy gradient (NSPG) that optimizes noisy latents using only clean action-space value estimates. Building on this result, we formulate a KL-regularized policy improvement over noisy latents and show that the resulting objective admits a diffusion-compatible regression form, avoiding backpropagation through the denoising process. Empirical results on state-based D4RL benchmarks and vision-based OGBench tasks demonstrate that the proposed noisy-space objective provides a principled and effective basis for training diffusion policies in offline reinforcement learning. Project webpage: https://mahmoud-selim.github.io/NSPG/

We study a two-server load balancing system with heterogeneous service rates that are a priori unknown to the dispatcher. The goal is to route customers according to the Shortest–Expected–Delay (SED) policy, but this requires knowledge of the service rates. Empirical policies that route based on estimates perform poorly: due to estimation error, the empirical policy disagrees with the oracle on an infinite region of the state space. We propose an online learning algorithm that converges to SED while learning the service rates. The algorithm carefully balances empirical SED routing with forced exploration phases that guarantee sufficient sampling of both servers. We prove that our algorithm achieves finite regret; this differs from classical Multi-Armed Bandit settings where regret typically grows logarithmically in time. Finally, numerical experiments demonstrate the performance of our algorithm and highlight the regimes in which forced exploration is especially beneficial.

Reasoning over language instructions in embodied tasks such as robotics often requires understanding spatial relations from a speaker's situated perspective. Humans infer such perspectives from shared environmental knowledge, activity context, and commonsense. Recent vision-language models (VLMs) appear capable of spatial reasoning, but their ability to infer a speaker's viewpoint from contextual cues and interpret situated spatial relations from that viewpoint remains unclear. We call this capability contextual observer grounding. To study this capability, we construct the Point-of-View Benchmark (POVBench), a dataset of 3D scenes and queries that disentangles Inferred, Stated, and Given forms of observer grounding in natural embodied communication. Given multi-view observations and a natural-language sentence, models must localize unseen or underspecified targets from situated spatial and contextual cues. Across state-of-the-art VLMs, localizing targets from directional language remains challenging, even when observer grounding is made explicit. We find that explicit breakdowns of observer-relative spatial reasoning improve target localization. Our project page is available at https://mimo-owl.github.io/POVBench/.

Does prolonged interaction with Large Language Models reduce users' observable autonomy during conversation? We test this degradation hypothesis by scoring user autonomy across 2,877 multi-turn conversations (46,952 turns) from WildChat and validating on 500 conversations from LMSYS-Chat-1M. Our core finding is a null result: population-level autonomy remains stable throughout conversations, only 4.7% of conversations show meaningful decline, and an apparent decrease at later turn positions is explained by survivorship bias. Instead, the dominant pattern is not erosion but absence: 80.8% of conversations exhibit zero or low autonomy throughout. When autonomy engagement occurs, it is self-reinforcing (persistence doubles from 40% to 85% over consecutive turns). These findings reframe the problem: the risk is not that LLMs progressively degrade users' observable autonomy, but that most users never engage in independent reasoning during the interaction.