让下一篇重要论文，不再错过。

Recent neural decoding approaches achieve strong performance by training on large collections of paired neural-behavioral recordings, but such datasets are expensive, invasive, and difficult to scale. In contrast, behavioral data is abundant and easy to collect, raising the question of whether neural decoding can benefit from decoupling behavioral and neural representation learning. We introduce BEEMO, a flexible multi-session framework that enables training with arbitrary mixtures of paired neural-behavioral recordings and unpaired behavioral data. A behavior decoder learns movement dynamics directly from behavioral trajectories, while neural activity is incorporated through cross-attention layers when available, allowing behavioral and neural supervision to scale independently. Across multiple intracortical motor datasets and tasks in nonhuman primates, we show that incorporating unpaired behavioral data improves decoding performance, particularly when transferring to new tasks, and that adding even a small amount of behavioral data from the target task further improves transfer. By decoupling neural and behavioral data, we unlock a new dimension for scaling: unpaired behavioral data, which is substantially cheaper and easier to collect than paired neural recordings, can be used directly to improve decoding performance. We further show that behavior-only pretraining, without any neural pretraining, achieves performance comparable to models pretrained on substantially larger neural datasets. Together, our results suggest that scaling behavioral data offers a practical and cost-effective path toward neural decoding models that generalize across subjects and tasks. Code is available at https://github.com/nerdslab/beemo.

We consider the problem of cross-validation (CV) in time series forecasting, with a focus on global forecasting models based on deep learning. The prevailing CV strategy in forecasting is to partition data across time with the last part of each series used for testing. With global models trained on collections of time series, however, a fundamentally different approach becomes possible: partitioning across series and time. We formalize series-wise CV for global forecasting models and benchmark several instantiations, including K-fold CV, bootstrapping, and Monte Carlo CV, on 11 datasets with 10 neural network architectures. We evaluate CV procedures along five dimensions: performance estimation, overoptimism rate, model selection accuracy, regret, and downstream forecasting performance. We withhold a random fraction of series from training so that they appear only in the test set. This protocol simulates the emergence of new series not seen during training, a common scenario in global forecasting. Under this protocol, our results reveal a nuanced trade-off: series-wise CV yields better downstream forecasting accuracy by guiding model selection toward configurations that generalize across both time and series, while time-wise CV retains a slight advantage for estimating out-of-sample error.

Diffusion-based molecular docking requires learning a corrective score field across a broad corruption distribution, yet individual pose errors can differ substantially in their structural consequences. This creates a post-training conflict between preserving broad score-field coverage and concentrating supervision on corrections that appear more relevant to docking geometry. Using DiffDock, a fixed-budget study compares broad continuation with torsional supervision focused on Cartesian consequence or conditional target consistency, and with a broad/focused mixture. Evaluation separates fixed score fidelity, generated-pose quality, best-of-10 sampling capacity, and confidence-selected Top-1 docking performance. At <5 Angstrom, broad continuation had the highest mean Top-1 success among trained policies, and focused or mixed supervision showed no consistent advantage under the tested budget. The starting model had better fixed score fidelity, whereas the evaluated broad-continuation model had higher mean Top-1 success with the same aggregate best-of-10 recovery. Focusing supervision on structurally consequential or more consistently specified corrections did not consistently improve returned-pose success over broad continuation.

When an image classifier decides, we can usually only estimate why. For quantum and all-optical classifiers that act linearly on amplitude-encoded pictures up to the measurement, with no trained classical network before or after them, each class's Born probability above its no-information share, $P_g - b_g$, splits exactly into contributions of the encoded amplitudes, shown as Born evidence maps (BEM). The split itself follows from known results on quadratic forms: with the transmitted light held fixed, it is what Integrated Gradients from a zero baseline and the Shapley value give. What we add is a way to measure it on the hardware that classifies, from class probabilities on a quantum chip or, in simulation, from camera frames on a lens bench, with no gradient circuit and no light sent backwards; a $g\times g$ map takes $1 + 2g^2$ runs. Deleting the 5% of pixels with the highest evidence drops accuracy from 73% to 9% (one seed), as occlusion and gradient $\times$ input do, while plain gradients remove much less. The classifiers share one architecture, a Fourier-plane mask followed by a brickwall circuit on qubits or by phase plates on a lens bench, read by a fixed Born headroom readout with no trained parameter; centring each picture on its own mean lets dark features count as much as bright ones. In simulation, four phase plates at one wavelength reach 72.3% on Caltech-10 colour photographs. On IBM's ibm_kingston, with the mask computed classically before approximate loading, 300 test pictures were classified at 37.7% against 10% chance (56.3% in exact simulation), and $2\times2$ maps of the filtered input of 30 test pictures drawn at random in advance, measured on the chip, agreed with the exact ones ($r = 0.88$; standard estimator 0.82). The maps also revealed a shortcut in the data; as the trim that removed it was chosen on test pictures, the Caltech-10 results are exploratory.

Reinforcement learning (RL) has become a standard post-training method for diffusion models, but most RL objectives roll out the current policy at every iteration, and with a many-step sampler these rollouts typically dominate the cost of training. Few-step models cut this cost, yet they model an average velocity over an interval, whereas existing RL objectives are built for an instantaneous velocity with a many-step sampler. We adapt RL objectives to few-step models systematically in three phases, namely a transformation, a decomposition and an adaptation. The transformation to $x$-space expresses each objective we consider through a common policy prediction, at the clean end by default and at an explicitly marked level for transition or flow-map objectives, so that objectives stated over densities, velocities or rewards become comparable. The decomposition writes the gradient of each of these objectives as the sum of a self-regression toward a reference and a reward augmentation, so that a method reduces, at the gradient level, to a choice of anchors and weights. The adaptation computes each anchor from a few-step model at the end of its transport or along its tangent, with the instantaneous velocity recovered from the network itself or from an auxiliary network. This yields few-step versions of FlowGRPO, AWM, DiffusionNFT, DiffusionOPSD and ReFL in one modular codebase. In our experiments, reward-gradient-based methods are the most effective when the reward is differentiable and its gradient is affordable; otherwise DiffusionNFT-style methods generally reach high rewards in relatively few iterations despite a sometimes higher per-iteration cost. Extra inference steps show no systematic aggregate benefit in our experiments, although selected outputs visibly improve with two to four times as many sampling steps as in training.

Self-attention carries a quadratic cost in token count, which limits how far Vision Transformers can scale with input resolution. We introduce SSoG, a separable sum-of-Gaussians spatial operator that replaces dense self-attention with a learned mixture of axis-factorized Gaussian kernels, reducing the per-layer cost from $\mathcal{O}(N^2)$ to subquadratic in $N$. Unlike dot-product self-attention, whose Lipschitz constant is unbounded outside a compact input domain, SSoG with fixed kernels is globally Lipschitz continuous, which, via the Picard--Lindelöf theorem, guarantees that the continuous-depth variant has a unique solution on the whole integration interval, a global guarantee that dense-attention counterparts lack in general. On CIFAR-100 at $224\times224$, in a matched discrete-Transformer setting and in the continuous setting, SSoG improves Top-1 accuracy by 2.2--4.1 points across three model widths while using 8.8--16.6\% fewer parameters, with depth and training recipe held fixed. In a looped-Transformer setting, SSoG improves Top-1 accuracy by 5.4--8.1 points. We further probe inference-time scaling by evaluating on inputs up to $1024\times1024$ at matched batch size: in the continuous-depth (Neural ODE) architecture, SSoG reaches up to 154\% higher throughput and over 60\% lower peak memory than dense attention, with the gap widening as resolution grows, consistent with the difference in asymptotic cost. In the discrete and looped settings, SSoG also achieves over 30\% higher throughput than dense attention at this resolution. Together, these results establish SSoG as a practical, parameter-efficient, and theoretically grounded alternative to dense attention across discrete, weight-shared, and continuous-depth Transformer variants.

Cryptocurrency transaction analysis often relies on black-box machine learning models that achieve high predictive performance but offer limited interpretability of transaction behaviors. This paper investigates the capabilities of large language models (LLMs) in analyzing real-world Bitcoin transaction graphs through a three-level framework: foundational metrics, characteristic overview, and contextual interpretation. To address token constraints, we introduce LLM4TG, a compact, human-readable graph representation, and CETraS, a connectivity-enhanced graph sampling algorithm. Experiments across five LLMs show that node-level information extraction achieves over 98.50% accuracy on most metrics, while meaningful graph characteristics are identified in up to 95.00% of responses. For address classification with limited labeled examples, top-1 accuracy reaches 50.49% using raw graphs, while top-3 accuracy reaches 72.43% using engineered features. These findings demonstrate the potential of LLMs for transaction graph analysis while revealing limitations in global graph reasoning, precise address classification, and explanation reliability.

In high-dimensional scientific measurements, many features in a single sample often capture the same underlying physical attributes, producing severe multicollinearity. Supervised feature selectors fit on the entire dataset, so a feature can be informative for classification overall while still redundant with another feature in the same class. Class-conditional selectors, by contrast, trade some accuracy for a narrower, per-class view, since they never see data outside one class, but this lets new classes be added without refitting the whole model. They still cannot determine, however, which feature to drop from a correlated group. To close this gap, we propose classconditional multicollinearity pruning (CCMP), which iteratively prunes one feature from every highly correlated pair, per class. Four theoretical guarantees establish explicit bounds on the error induced by pruning under stated assumptions. Across four benchmark datasets spanning spectroscopic and hyperspectral imaging modalities, CCMP retains as few as 42 of 1531 features (3%) and as many as 60 of 145 (41%), outperforms every classconditional baseline, and remains competitive with supervised methods.

Proximal Policy Optimization (PPO) is widely used in reinforcement learning due to its strong empirical performance, yet it lacks formal guarantees for policy improvement and convergence. PPO's clipped surrogate objective is motivated by a lower bound on linearization of the value function in flat geometry setting. We derive a tighter surrogate objective and introduce Fisher-Rao PPO (FR-PPO) by leveraging the Fisher-Rao (FR) geometry. Our scheme provides strong theoretical guarantees, including monotonic policy improvement. In the direct parametrization setting, we show that FR-PPO achieves sub-linear convergence, and for parametrized policies we further obtain sub-linear convergence up to the compatible function approximation error. Finally, although our primary focus is theoretical, we also demonstrate empirically that FR-PPO can be implemented as a drop-in replacement for PPO and achieves performance comparable to PPO across a range of standard reinforcement learning tasks.

Language-model agent systems must map natural-language requests to one or more specialized agents before planning or tool execution. This decision is often treated as top-1 selection, although compositional requests can require several capabilities and unnecessary dispatch raises cost. We formulate fixed-catalog routing as set-valued language understanding and construct a benchmark of 3,000 WildChat prompts over 12 capability-defined agents. Reproducible schema-guided reference sets are balanced by agent frequency and route cardinality. We compare profile retrieval, multilabel classifiers, a fine-tuned text encoder, and catalog-conditioned zero-shot GPT-4o. Evaluation combines set quality, latency, capability coverage, and ordinal-cost utility. The encoder attains 89.59 F1, compared with 71.79 for linear SVM and 41.51 for GPT-4o, and retains 88.02 F1 under a harder semantic-group split. We further introduce Weighted Agent Routing (WAR), a deterministic post-scoring rule that changes the quality-cost operating point without retraining; Encoder+WAR obtains the best constrained utility. The results show that catalog-specific supervision and explicit set construction are important for multi-agent dispatch, while cost-aware decision layers can complement strong language encoders.

Recursive training of generative models on their own outputs can lead to model collapse, a compounding drift away from the true data distribution. Existing theoretical works bound finite-round error accumulation in the context of diffusion models, but two questions remain open:~what distribution does the recursion converge to, and how fast? We answer both, isolating a mechanism distinct from imperfect learning: even with perfect score estimation and exact sampling, the early stopping of the reverse diffusion (required for numerical stability) drives a progressive drift away from the data distribution. We prove that this recursion converges geometrically to a unique limiting distribution, which admits a closed-form characterization as an infinite mixture of increasingly Gaussian-smoothed versions of the data distribution. A Hermite spectral decomposition of this limit reveals that recursive training acts as a low-pass filter: higher-order modes, which encode fine non-Gaussian structure, are attenuated much more strongly than coarse modes. This spectral picture motivates annealed truncation schedules that progressively shrink truncation times across retraining rounds; we prove that any schedule converging to $0$ asymptotically eliminates recursive compounding. Finally, we show our idealized characterization is robust: in the presence of discretization and score estimation errors, the learned distribution remains in a Wasserstein-2 ball around the ideal limit, with mode-dependent contraction rates that contract high-order errors faster than low-order ones. We validate the theory on synthetic Gaussian mixtures and CIFAR-10.

Many complex networks exhibit hierarchical, tree-like structures, making hyperbolic space a natural candidate wherein to learn representations of them. Based on this observation, Hyperbolic Graph Neural Networks (HGNNs) have been widely adopted as a principled choice for representation learning on tree-like graphs. In this work, we question this paradigm by proposing the additional condition of geometry--task alignment, i.e., whether the metric structure of the target follows that of the input graph. We start by empirically showing that HGNNs can recover low-distortion representations when supervision explicitly rewards metric preservation and proceed by theoretically analyzing a controlled regression setting in which a linear readout benefits from representations whose geometry is aligned with the target metric. Finally, by relying on both predictive performance and embedding distortion, we further show that HGNNs gain an advantage on link prediction, a naturally geometry-aligned task, whereas this advantage largely disappears on standard node classification benchmarks. Overall, our findings shift the focus from only asking "Is the graph hyperbolic?" to also questioning "Is the task aligned with hyperbolic geometry?", showing that HGNNs consistently outperform Euclidean models under such alignment, while their advantage vanishes otherwise.

sEMG Foundation Models aim to learn general purpose representations over diverse subjects and recording conditions that boost performance on a variety of downstream tasks. However, existing evaluations make it difficult to discern whether pretraining alone is sufficient to combat the cross-subject overfitting which makes machine learning on sEMG so challenging. We conduct a systematic evaluation of three sEMG foundation models across three gesture classification datasets under both intra-session and inter-subject splits, using three fine-tuning strategies designed to isolate the contribution of pretraining. We find that pretraining boosts performance when train and test samples are drawn from the same recording session, but this benefit largely disappears under inter-subject evaluation. In low subject count datasets, pretraining provides little to no benefit and classical ML baselines match or exceed foundation model performance.

The use of algorithmic predictions in decision-making leads to a feedback loop where the models we deploy actively influence the data distributions we see, and later use to retrain on. This dynamic was formalized by Perdomo et al. 2020 in their work on performative prediction. Our main result is an unconditional reduction showing that any no-regret algorithm deployed in performative settings converges to a (mixed) performatively stable equilibrium: a solution in which models actively shape data distributions in ways that their own predictions look optimal in hindsight. Prior to our work, all positive results in this area imposed strong restrictions on how models influenced distributions. By using a martingale argument and allowing randomization, we avoid any assumption on how populations respond to predictions and sidestep recent hardness results showing that deterministic stable models are in general PPAD-hard to compute. Lastly, on a more conceptual note, our connection sheds light on why common algorithms, like gradient descent, are naturally stabilizing and prevent runaway feedback loops. We hope our work enables future technical transfer of ideas between online optimization and performativity.

We identify an intriguing phenomenon when linearly interpolating between permutation-aligned neural networks: test accuracy briefly improves near the interpolation endpoints before dropping toward the midpoint, which we term BUMP (Brief Unexpected Mid-interpolation Peak). We observe that BUMP appears consistently across datasets, architectures, training regimes, and remains robust under substantial perturbations such as early checkpoints and extreme pruning. Moreover, controlled ablations suggest that it cannot be attributed to trivial regularization effects. While similar near-endpoint gains are visible in prior interpolation plots, they have not been explicitly documented or systematically and thoroughly analyzed. Leveraging this effect, we propose a simple, novel ensembling scheme: weight-prediction mixed ensembling that combines weight mixing with prediction-space mixing. Across our experiments, the resulting ensembling method consistently outperforms standard logit ensembling in test accuracy and yields improved calibration trade-off, highlighting complementary benefits of weight- and prediction-space mixing.

Exploration has been a focus of reinforcement learning research for a long time. Recently, there has been growing evidence that it is also an important ingredient in LLM reinforcement learning recipes that can significantly impact downstream performance. Many existing methods control exploration in the action-space, for example, using temperature scaling. However, these methods cannot reorder tokens but only influence the variance in the output distribution. This limits exploration and can lead to divergence or stalled training. Here, we investigate parameter-space exploration, where rollouts are generated by sampling different policies from a posterior that may each explore different rollouts. Sampling less or more diverse policies is then a complementary control lever over exploration. We introduce a family of methods called Perturbed Parameter Policy Optimization (3PO) which use different sampling strategies and different rollout grouping for reward estimation. Experiments on Olmo-3-1025-7B and Qwen2.5-Math-7B across mathematical reasoning and code generation tasks show that these approaches can improve the average downstream performance over standard GRPO at a near-identical FLOPs cost. Moreover, using multiple parameter samples generally produces fewer zero-advantage groups and malformed or incorrect rollouts during training than GRPO and action-space baselines. Overall, our work presents evidence that parameter-space exploration can improve reinforcement learning for LLMs.

Socially interactive humanoids aim to interact with humans through their bodies, adapting in real time to a partner's movement, intent, and abilities. This requires models that understand both how bodies move and what their movements mean. Yet, existing evaluation frameworks for interactive motion generation do not assess whether generated reactive motions are valid or appropriate within a shared movement vocabulary. Existing frameworks rely on kinematic metrics such as FID and beat alignment that cannot capture this semantic property, while improvised interaction datasets lack fine-grained move annotations. Inspired by computational evaluations of games such as Go and Chess, we propose Salsa as an embodied interaction evaluation domain. Salsa is interactive, dyadic, and governed by a move vocabulary and community-validated judging criteria covering timing, musicality, technique, difficulty, partnering, and originality. We present CoMPAS3D, a motion capture dataset of improvised partner salsa dance and an evaluation framework covering kinematic quality, two objective metrics (move legibility and proficiency appropriateness), and six competition-based subjective dimensions. The dataset comprises 3 hours of improvisation by 18 dancers spanning beginner, intermediate, and professional skill levels, with over 2,800 expert-annotated segments covering move types, execution errors, and stylistic elements. We define three benchmark tasks: move classification, proficiency estimation, and follower generation. Fine-tuned vision-language models perform well on ground-truth sequences and correlate with subjective expert judgments, suggesting their promise as measurement tools. Applied to Duolando and InterGen, our objective metrics reveal a large gap between ground-truth and generated dances, corroborated by expert human evaluations across six judging dimensions.

The Rectified Power Unit (RePU) activation function, a differentiable generalization of the Rectified Linear Unit (ReLU), has shown promise in constructing neural networks due to its smoothness properties. However, deep RePU networks often suffer from critical issues such as vanishing or exploding values during training, rendering them unstable regardless of hyperparameter initialization. Leveraging the perspective of effective field theory, we identify the root causes of these failures and propose the Modified Rectified Power Unit (MRePU) activation function. MRePU addresses RePU's limitations while preserving its advantages, such as differentiability and universal approximation properties. Theoretical analysis demonstrates that MRePU satisfies the criticality conditions in the limit of a small kernel, and a finite-kernel analysis yields an initialization rule for deep MRePU networks. Experiments on polynomial regression, physics-informed neural networks (PINNs) and real-world vision tasks show that MRePU networks initialized by this rule remain trainable at depths at which RePU networks fail, while their accuracy relative to standard activations depends on the task.

In this note we prove that the Langevin Monte Carlo (LMC) algorithm can minimize the population risk of appropriately regularized depth-2 nets with no restrictions on the size of the net and for any choice of training data --- and we give non-asymptotic convergence rates for it. We achieve this via showing that in q-Renyi divergence, the iterates of LMC converge to the Gibbs distribution of certain Frobenius norm regularized losses for these nets --- in both classification and regression settings. The key technicality of the work is two-fold (a) to keep the amount of regularization needed be independent of the size of the net and (b) to obtain convergence rates that scale with just the Poincare constant of the corresponding Gibbs measure. This work synthesizes several recent observations about isoperimetry conditions under which LMC converges and that two-layer neural loss functions can always be regularized by a constant amount such that they satisfy the Villani conditions, and thus their Gibbs measures satisfy a Poincare inequality.

Bilateral trade models the task of intermediating between two strategic agents, a seller and a buyer, willing to trade a good for which they hold private valuations. We study this problem from the perspective of a profit-maximizing broker in a regret minimization framework. At each time step, a new seller and buyer arrive and the broker proposes a trading price to each agent. The trade happens if both agents accept, and the profit of the broker is the difference between the price received from the buyer and the one paid to the seller. At the end of each interaction, the learner only observes whether the trade occurred, and its goal is to achieve sublinear regret relative to the best fixed pair of prices in hindsight. We start by investigating the adversarial version of the problem, where buyer and seller valuations are generated by an adversary. In this setting, we construct a polynomial-time learning algorithm with a $\tilde O(T^{\frac 23})$ regret rate, which is nearly tight even if the valuations are drawn i.i.d. from independent distributions, thus matching the rate for one-sided pricing. Our adversarial algorithm combines a localized estimation procedure with an information-based minimax construction. We then consider valuations drawn i.i.d. from independent distributions satisfying the monotone-hazard-rate (MHR) property and provide a learning algorithm with nearly tight $\tilde{O}(\sqrt{T})$ regret. The algorithm combines localized estimation with a stochastic convex bandit routine applied to the negative logarithm of expected profit. Finally, we complement this result by showing a separation between MHR and regular distributions, proving that the minimax regret rate in the latter setting is $\tilde Θ(T^{\frac 35})$. This last result contrasts with the $\tilde{O}(\sqrt{T})$ regret bound that can be obtained for one-sided pricing with regular distributions.

Despite the numerous applications of convex constraints in Robotics, enforcing them within learning-based frameworks remains an open challenge. Existing techniques either fail to guarantee satisfaction at all times, or incur prohibitive computational costs. This paper presents RAYEN, a framework for imposing hard convex constraints on the output or latent variables of a neural network. RAYEN guarantees constraint satisfaction during both training and testing, for any input and any network weights. Unlike prior approaches, RAYEN avoids computationally expensive orthogonal projections, soft constraints, conservative approximations of the feasible set, and slow iterative corrections. RAYEN supports any combination of linear, convex quadratic, second-order cone (SOC), and linear matrix inequality (LMI) constraints, with negligible overhead compared to unconstrained networks. For instance, it imposes 1K quadratic constraints on a 1K-dimensional variable with only 8 ms of overhead compared to a network that does not enforce these constraints. An LMI constraint with 300x300 dense matrices on a 10K-dimensional variable can be guaranteed with only 12 ms additional overhead. When used in neural networks that approximate the solution of constrained trajectory optimization problems, RAYEN runs 20 to 7468 times faster than state-of-the-art algorithms, while guaranteeing constraint satisfaction at all times and achieving a near-optimal cost. Finally, we demonstrate RAYEN's ability to enforce actuator constraints on a learned locomotion policy by validating constraint satisfaction in both simulation and real-world experiments on a quadruped robot. The code is available at https://github.com/leggedrobotics/rayen

LLMs encode both general capabilities and domain-specific knowledge in a single set of parameters. We ask whether this capacity can be reorganized: keeping broadly useful computation in a shared backbone, while moving specialized knowledge into external memory modules. We propose \emph{knowledge offloading} (KOFF), a framework for decomposing a pretrained LLM into a sparse shared backbone and domain-specific memories. Starting from a frozen base model, we either jointly or sequentially learn a structured pruning mask and lightweight recovery modules, implemented as LoRA adapters and learned key-value caches. Across Llama and Qwen models from 3B to 14B, we find that 12\% to 20\% of capacity can be moved out of the shared backbone while largely preserving perplexity and MMLU performance (up to 90\% of the full model). We show that LoRA and learned KV memories are complementary, and analyses suggest that the offloaded knowledge is specialized: mismatched memories do not yield the same gains, and language-specific neurons are preferentially removed while language-neutral neurons largely remain in the backbone. These results suggest that the knowledge of a pretrained LLM can be reallocated between a shared core and swappable external memories.

Attention-based time series forecasters commonly group consecutive observations into patches. Patch length is known to have a substantial impact on forecasting performance, yet its selection remains largely empirical. We develop a theoretical framework that explains why patch length matters for forecasting accuracy and provides a principled basis for its selection. For linear state-space models, we derive an exact factorization of the forecasting coefficient matrix that separates forecast dynamics from filtering dynamics. We show that dominance of a single forecast component at long horizons does not generally imply a rank-one representation. This explains why longer horizons alone need not reduce representational complexity. We combine the factorization with estimation bounds to develop a patch-selection criterion that balances representation and estimation errors under temporal dependence. Our analysis establishes conditions under which balanced patching is justified and clarifies how predictive structure, head count, training sample size, and long-run dependence enter the selection criterion.

We provide a rigorous random matrix theory analysis of spiked cross-covariance models where the signals across two high-dimensional data channels are partially aligned. These models are motivated by multi-modal learning and form the standard generative setting underlying Partial Least Squares (PLS), a widely used yet theoretically underdeveloped method. We show that the leading singular values of the sample cross-covariance matrix undergo a Baik-Ben Arous-Peche (BBP)-type phase transition, and we characterize the precise thresholds for the emergence of informative components. Our results yield the first sharp asymptotic description of the signal recovery capabilities of PLS in this setting, revealing a fundamental performance gap between PLS and the Bayes-optimal estimator. In particular, we identify the SNR and correlation regimes where PLS fails to recover any signal, despite detectability being possible in principle. These findings clarify the theoretical limits of PLS and provide guidance for the design of reliable multi-modal inference methods in high dimensions.

Minimal filling architectures describe the smallest hidden-layer widths for which the neurovariety of a polynomial neural network fills its ambient function space. It was conjectured that every such architecture has unimodal hidden widths. We construct counterexamples to this conjecture using a frontier search and certify the main example through finite-field Jacobian computations and recursive dimension bounds. In the opposite direction, we prove that every minimal filling architecture of depth at most three is unimodal. A nonunimodal minimal filling architecture of depth four shows that the depth bound is sharp.

Image classification models often encounter objects in diverse and unpredictable contexts at test time, which may differ substantially from those seen during training. This phenomenon, termed context shift, exposes a key fragility in neural networks: when models rely on non-causal contextual cues for prediction, shifts in context can lead to significant performance degradation. In this paper, we propose a novel training framework to improve the robustness of convolutional neural networks under context shift. Our approach leverages spatial feature attribution to guide models toward making predictions that rely less on contextual regions and more on object-relevant features. As a first step, recognizing a theoretical limitation of existing feature attribution methods we introduce a new variant, ContrastiveCAMs, which produces more faithful attribution maps of model predictions. Building on ContrastiveCAMs, we further propose Context-Regularized Cross-Entropy (CR-CE), a modification of the standard cross-entropy loss that regularizes the model's attention by suppressing the influence of contextual regions, thereby improving robustness to context shift. We evaluate the effectiveness of our approach on several medium-to-large scale datasets (Waterbirds, Spawrious, ImageNet/ImageNet-BG, Hard-ImageNet), and report consistent improvements in context-shift robustness.

World-action models adapt pretrained video generators, and vision-language-action models build on pretrained vision-language backbones; both tie policy architecture and inference cost to large transformers designed for generation rather than control. It remains unclear how much these generation-oriented choices affect control. In this work, we introduce Latent Action JEPA (La-JEPA), a compact world-action model whose policy transformer is trained from scratch in the feature space of a pretrained V-JEPA 2.1 encoder. Following Fast-WAM, La-JEPA learns to predict future features jointly with actions during training and generates only actions at inference. Among policies without embodied pretraining, La-JEPA achieves state-of-the-art performance in clean and randomized RoboTwin 2.0 scenes and on LIBERO-Plus with correct instructions, while using far fewer parameters than the strongest competing policies. It runs substantially faster than Fast-WAM at comparable performance and outperforms the closest JEPA-based methods, LeapBot-WA and JEPA-WAM, on LIBERO-Plus and RoboTwin 2.0 without their pretrained policy backbones. On four real-robot tasks, its average success is comparable to pi0.5 and above non-embodied-pretrained baselines. Through a design study of the choices behind La-JEPA, we show that JEPA features outperform video VAE latents, most clearly out of distribution, and that training-time future prediction choices significantly impact the policy even when future prediction is skipped at inference. We correct the LIBERO-Plus instructions, as the officially released evaluator corrupts them for most variants, and re-evaluate several strong open baselines. Without a generative pretrained policy backbone, La-JEPA is better and faster than far larger policies, suggesting that robot foundation models may benefit from architectures designed for control rather than inherited from generation.

Semantic communication (SemCom) using a relay node is considered, where the relay assists in conveying task-relevant information from a source to a destination, with meaning defined by classification labels. A joint end-to-end design of the source encoder, relay transformation, and destination decoders is developed to optimize both semantic inference accuracy and perceptual reconstruction quality for image data. Relay-assisted schemes are shown to outperform direct transmission, with gains that become more pronounced in low signal-to-noise ratio (SNR) regimes across different relaying strategies. Alongside these performance benefits, however, a key security vulnerability introduced by the relay is identified, and a Byzantine attack is formulated in which the relay injects structured, learned perturbations into its transmitted signal. The attack operates on intermediate communication representations subject to channel effects, with perturbations generated through a learned mapping applied at test time. It is shown to significantly degrade semantic inference accuracy at the destination. These results reveal a clear trade-off between SemCom gains and security risks in relay-assisted systems, motivating the need for secure relay designs that account for insider threats.

This study evaluates the added prognostic value of day-1 multimodal MRI for predicting three-month functional outcome after acute ischemic stroke (AIS). Sixty-six AIS patients with ADC, DWI, FLAIR, and clinical data were analyzed. Three-dimensional pretrained ResNet embeddings were extracted from each MRI modality, reduced using fold-isolated principal component analysis, and combined with structured clinical variables and ADC-derived lesion biomarkers. Linear support vector machines were evaluated using eight-fold stratified group cross-validation across MRI modality combinations. Clinical-only modeling achieved strong performance (AUC = $0.899 \pm 0.151$), while MRI-only models showed limited discrimination (AUC $\approx$ $0.67$-$0.73$). Among multimodal configurations, FLAIR-clinical fusion achieved the highest mean cross-validated performance (AUC = $0.948 \pm 0.061$), followed by ADC-clinical fusion (AUC = $0.941 \pm 0.094$). Full ADC-DWI-FLAIR fusion did not consistently outperform single-modality clinical fusion. Feature-importance analysis identified age, NIHSS subscores, ADC lesion burden, and learned FLAIR representations as key contributors, while qualitative FLAIR activation maps showed responses over lesion-like regions. These findings suggest that compact day-1 FLAIR and ADC representations provide complementary prognostic information beyond clinical variables and support interpretable, data-efficient outcome prediction in AIS.

We introduce Hyper Input Convex Neural Networks (HyCNNs), a novel neural network architecture designed for learning convex functions. HyCNNs combine the principles of Maxout networks with input convex neural networks (ICNNs) to create a neural network that is always convex in the input, theoretically capable of leveraging depth, and performs reliably when trained at scale compared to ICNNs. Concretely, we prove that HyCNNs require exponentially fewer parameters than ICNNs to approximate quadratic functions up to a given precision. Throughout a series of synthetic experiments, we demonstrate that HyCNNs outperform existing ICNNs and MLPs in terms of predictive performance for convex regression and interpolation tasks. We further apply HyCNNs to learn high-dimensional optimal transport maps for synthetic examples and for single-cell data, where they oftentimes outperform ICNN-based neural optimal transport methods and other baselines across a wide range of setting.