Papers

Hyperspectral imaging distinguishes RGB-similar objects, but a pretrained vision-language-action model needs a state its frozen decoder can use. SPECTRA-VLA provides a native-width spectral upgrade without retraining the visual encoder or action decoder. A learned continuous residual enriches each target token through its own normalized false-color reference; projection, norm control and validation-calibrated injection regulate the change. Across 86 held-out acquisitions, category balanced accuracy rises from 0.812 to 0.883. With identical residuals and scale, own-token addition improves reuse of an external category-derived probe on frozen decoder logits: 0.701 versus 0.590 with pooled replacement, at similar image-disjoint recognition. Five grouped tests covering the same 86 acquisitions recover within-category spectral shape with $R^2=0.802/0.337$ from continuous and posterior-code states. Category-matched spectral swaps further show that own-token composition approximately doubles the target-specific recovery increment after frozen decoding. Spectral enrichment and receiving-token preservation thus work together at the pretrained interface.

Understanding the security posture of cyber physical systems (CPS) requires reasoning across both information technology (IT) and operational technology (OT). However, a security analyst often holds an incomplete, role-specific view of the underlying system which has most likely resulted from their knowledge limited by a need-to-know framework. Such an incomplete view can potentially lead them to incorrectly assess if a cyber attack produces an untoward physical consequence resulting in making uninformed decisions about system resilience. In this work, we present AWARE-RG (Abstraction Aware Explanation of Resiliency Graphs), a semi-automated, human-in-the-loop framework that provides an analyst with a role-bounded, need-to-know explanations of resiliency paths in a CPS. The AWARE-RG abstraction structure is derived using Formal Concept Analysis over MITRE ATT&CK for ICS tactics, ordered by the ICS Cyber Kill Chain. For valid attack-fault queries, plan repair and backward means-end analysis identify the missing dependencies, explain why the physical hazard is reachable, and suggest defensive strategies. For invalid attack-fault queries, AWARE-RG localizes the first abstraction level at which the path fails, and reports confirmed blocking conditions as hardening pointers.

We conducted an investigation into the possible correlation between very-high-energy (VHE, $E>$ 0.1 TeV) gamma-ray sources, as identified by the Large High Altitude Air Shower Observatory (LHAASO), and high-energy neutrino events captured by the IceCube Neutrino Observatory. For this analysis, we utilized the First LHAASO catalog, which comprises 90 VHE gamma-ray sources, and the IceCube Event Catalog of Alert Tracks (IceCat-1), which includes 348 track-like neutrino events. Through Monte Carlo simulations, we assessed the correlations between the neutrino events and LHAASO sources, achieving the best p-values of 0.237, 0.312, and 0.115, respectively, for all LHAASO sources, KM2A sources, and WCDA sources. These findings imply that the spatial coincidence between LHAASO sources and neutrino alerts is relatively weak. Furthermore, under the assumption that gamma-rays are exclusively generated through hadronic interactions, we estimated the most optimistic theoretical neutrino fluxes from LHAASO sources and compared them with the IceCube observed fluxes. The results indicate that the majority of these sources are improbable counterparts to the detected neutrinos. In conclusion, LHAASO sources could be potential galactic neutrino emitters but currently do not exhibit compelling evidence, thereby underscoring the necessity for elevated sensitivity levels in forthcoming detection instruments like IceCube-Gen2.

We study the sharp Fourier extension inequality from $L^2(\mathbb S^{d-1})$ to the radial–angular mixed space $L^q_{\mathrm{rad}}L^2_{\mathrm{ang}}$. Our work is motivated by and builds upon the foundational analysis of Carneiro, Oliveira e Silva and Sousa \cite{COS}, who initiated the sharp mixed-norm spherical restriction problem from the viewpoint of spherical-harmonic decomposition. Their reduction revealed that the extremal problem is governed by a competition among the individual harmonic modes, and their results established degree-zero extremizers in several important ranges of exponents, including even integer exponents and the endpoint case. The present paper addresses this question by establishing a complete modal ordering result. Building on their spherical-harmonic reduction, the sharp-constant problem is reduced to the weighted Bessel moments \[ I_{\nu,k}(q) := \int_0^\infty |J_{\nu+k}(r)|^q r^{2\nu+1-\nu q}\,dr, \qquad \nu:=\frac d2-1. \] For every integer $d\ge2$ and every finite admissible exponent $q>2d/(d-1)$, we prove \[ I_{\nu,k}(q)<I_{\nu,0}(q), \qquad k\ge1, \] and hence constants are the unique nonzero extremizers of the sharp mixed-norm extension inequality. In dimensions $d\ge3$ we prove the stronger strict ordering \[ I_{\nu,0}(q)>I_{\nu,1}(q)>I_{\nu,2}(q)>\cdots. \] We resolve this difficulty by combining exact Hankel asymptotics with a finite remainder comparison. The modal ordering also yields the optimal quadratic global stability inequality. For every finite admissible $q$, \[ C_{d,q}^2\|f\|_2^2 -\|\widehat{f\sigma}\|_{L^q_{\mathrm{rad}}L^2_{\mathrm{ang}}}^2 \ge \Gamma_{d,q}\|f-P_0f\|_2^2, \] with the best possible coefficient. When $d\ge3$, this coefficient is explicitly \[ \Gamma_{d,q} =(2\pi)^d \left[ I_{\nu,0}(q)^{2/q}-I_{\nu,1}(q)^{2/q} \right]. \]

Accurately delineating Arctic coastal polynyas from sea ice concentration data remains challenging because of complex coastal geometry and spatially heterogeneous ice conditions. We develop a daily sea ice concentration (SIC) mapping framework based on hierarchical partitioning, connected-component analysis, and differentiated region growing for polynya and open-water regions. Applied to September-May ice seasons from 2002/03 to 2024/25, the framework achieves a mean overall accuracy of 96.74% and a mean kappa coefficient of 0.93 in the 63-scene visual assessment. Sensitivity experiments reveal a consistent pattern: the absolute magnitude of SIC-defined polynya extent is strongly configuration-dependent, whereas major temporal features are comparatively more stable. Relative to a 75% upper SIC threshold, seasonal extent integrals differ by 13.48%, 32.47%, and 65.04% at thresholds of 80%, 85%, and 90%, respectively, while the broad interannual structure remains recognizable. The response is also strongly regional and seasonal, with the largest threshold sensitivity in the Central Arctic Ocean, very weak sensitivity in the Gulf of St. Lawrence, and greater pan-Arctic curve separation from January onward. At an 80% threshold, product-comparison MAPE is 68.19% between the 6.25 and 25 km products and 11.47% between the 12.5 and 25 km products under their respective comparison periods and reference definitions. Regional product disagreement is likewise heterogeneous. Together, these results show that reproducible long-term polynya monitoring requires a fixed and explicitly documented combination of SIC product, threshold, and spatial classification rules, while broad temporal patterns can remain informative even when absolute extent differs substantially.

We study two-player zero-sum stochastic games in which a stream of stage payoffs is evaluated by a translation-invariant charge on the set of stages. Every patient valuation, that is, every diffuse charge that obeys the time value of money principle, is translation-invariant, and by Mertens and Neyman each player has, for every $\varepsilon>0$, a strategy that is $\varepsilon$-optimal for all patient valuations simultaneously. We show that this robustness breaks down for translation-invariant charges. In the Big Match with $0$–$1$ payoffs, whose uniform value is $1/2$, every strategy of Player 1 is worthless: for every strategy of Player 1 and every $\delta>0$, Player 2 has a pure reply, and there is a translation-invariant charge under which Player 1's payoff is less than $\delta$. Consequently, the Big Match has no value with respect to all translation-invariant charges. This contrasts with finite Markov decision processes, where every stationary Blackwell optimal strategy is optimal for every translation-invariant charge. We pose questions on robust guarantees in general stochastic games and on optimality with respect to a fixed translation-invariant charge.

AI workloads increasingly employ low-precision formats such as BF16 / FP16, FP8 variants, FP4, Micro Scaling (MX), and INT modes across models, layers, and execution stages to balance accuracy, memory footprint, and throughput, which in turn requires hardware to support these formats in a flexible and efficient way. We present MFSA, a weight-stationary multi-format systolic array that achieves area efficiency through two architectural choices. At the PE level, MFSA uses dot-product-accumulator-based processing elements extended from an existing open-source trans-precision FPU to support reduced-precision formats (FP32/FP16/BF16/CFP8/FP4/INT16/INT8/INT4) with a shared datapath and proportional throughput scaling. At the array level, MFSA employs Online Block-Scale Rebasing (OBSR) for block-scaled formats, which folds each block scale into the running FP32 partial sum at block boundaries without adding a second cross-block accumulator, thereby reducing the additional overhead of block-scaling support. In TSMC 28 nm at 1 GHz, matched-throughput BF16/FP8/FP4 cascaded FMA-PE baselines are 1.5x/2.5x/4.5x larger than the single-format DPA-PE, the gap growing with subword parallelism. A routed 16x16 MFSA reaches a peak 0.5/1/2/4 TOPS of FP32/BF16/FP8/FP4, and OBSR cuts the incremental MX overhead by 64.2% in area and 67.5% in energy versus a naive MX implementation baseline, without hurting the numerical stability on real LLM traces. Adding support for all other formats based on a BF16+CFP8 systolic array costs 25.7% additional area. OBSR also extends to non-power-of-two block scales: folding the fractional part of an E4M3 scale into the MFPE dot product supports NVFP4 alongside OCP MX for 11.8% area and 7.3% power over a CFP8+FP4 base, less than half the cost of a naive implementation.

Typed decision models need an interface that can use additional reasoning while retaining candidate outputs. We train a shared candidate-answer reader on a frozen Qwen3.5-4B backbone to unify direct decisions with on-demand, interruptible reasoning. The reader returns distributions from empty and partial reasoning prefixes. A benefit gate starts reasoning, a natural-state trigger requests explicit reads, and an entropy rule accepts acquired answers and cancels generation. Measurements distinguish answer-query work, generated output, and pipeline time. Across 768 questions from ProofWriter, LogiQA 2.0, and ARC-Challenge, accuracy rises from 70.96% for direct reading to 84.38%. Relative to 512-token terminal reading, accuracy improves by 7.29 percentage points with 2.25% more mean output work. Relative to 1,024-token terminal reading, pipeline time falls by 46.30% with an accuracy difference of -1.17 points. Exact-prefix controls isolate the saving from interruption, while read-scheduling controls distinguish reduced observation work from shorter response time.

Individualized survival predictions are often summarized by scalar risk scores or event probabilities at selected time horizons, although clinically relevant risk may also differ in its temporal organization. We consider finite-dimensional representations of individual hazard trajectories through interpretable hazard functionals, such as peak timing, cumulative hazard over prespecified intervals, and persistence of elevated risk. A central issue is that the richness of such representations is constrained by the hazard geometry supported by the underlying survival model. We formalize this through model-induced tangent spaces and functional ranks, derive an exact decomposition of model-supported variation into directions retained and lost by a chosen representation, and introduce a model-relative notion of local completeness. We further distinguish intrinsic model capacity from the variation actually activated by covariates. These results motivate a structured extension of AFT time scaling through a positive-affine deformation on the log-time scale, yielding a power-warp mechanism with an additional temporal deformation direction. Combined with a log-logistic reference hazard, this gives a parsimonious likelihood-based working specification for which peak timing and peak height provide complete coordinates over the unimodal regime. We also establish plug-in recovery and asymptotic inference for hazard functionals. Simulations and a lung-cancer application illustrate how model-induced restrictions propagate to hazard-functional representations.

Reinforcement learning provides a direct mechanism for optimizing task objectives in large language model-driven agents, but task rewards are typically sparse trajectory-level signals that only evaluate task success, without directly supervising action-conditioned environment transitions, leaving the model with limited explicit understanding of action consequences and environment dynamics. Recent work introduces auxiliary supervision through policy and world modeling co-training, yet typically treats complete environmental observations uniformly as world modeling targets, overlooking the varying decision relevance of observation content. We propose DGS-WM (Decision-Guided Selective World Modeling), a decision-guided policy-world modeling co-training approach for language agents. DGS-WM employs a Decision-Guided Observation Selection component that filters observation units with decision value as world modeling targets based on their support for subsequent policy responses, and a Confidence-Adaptive Weighting component that adaptively adjusts supervision strength according to the model's prediction confidence on selected targets. Together, these components form a decision-aware auxiliary world modeling loss, jointly optimized with the RL loss on the same policy model. Across multiple benchmarks covering single-hop/multi-hop question answering and interactive web tasks, DGS-WM achieves consistent improvements under different model scales and RL algorithms, demonstrating more stable and effective performance compared to full-observation world modeling methods.

Community-authored augmented reality (AR) can contest dominant heritage narratives by juxtaposing stories with physical surroundings. We examine the Thámien Ohlone AR Tour, developed with members of the Ohlone Native American community at Mission Santa Clara and Santa Clara University. Combining paired pre/post surveys from 57 participants with collaborative thematic analysis, we compare reception among Ohlone community members, university affiliates, and Mission visitors. Spatial dissonance captures felt tensions when AR challenged familiar campus readings or when surroundings conflicted with Ohlone histories the AR affirmed, such as a garden's significance as a burial ground. Cultural frames shaped interpretation; contemporary people and relationships made Ohlone continuance tangible. Participants appreciated the tour and called for permanent physical acknowledgment. We propose staging reflective encounters, clarifying community-authored meanings, and using AR to prototype public acknowledgment and support community-directed advocacy. These findings foreground relationships to land and place, informed by Indigenous scholarship, in designing and evaluating AR.

The scarcity of observed biodegradation reactions constrains the training of product and pathway predictors. We present enviGraph, a framework that generates and curates synthetic reaction data, uses these data to pretrain a graph-to-sequence predictor, and fine-tunes the predictor on observed environmental reactions for product and recursive pathway prediction. Applying inherited environmental transformation rules to structures from PubChem, NORMAN SusDat and EPA DSSTox, followed by curation, produces 776,851 canonical parent-product pairs for pretraining. We compare this dataset with three United States Patent and Trademark Office (USPTO) reaction datasets: USPTO-50K, USPTO-MIT and USPTO-Stereo. We evaluate enviGraph and the sequence-based enviFormer using ten folds split by pathway on the EAWAG Biocatalysis/Biodegradation Database (BBD), regulatory pesticide biodegradation data (Soil) and activated-sludge biodegradation data (Sludge). For both predictors, synthetic pretraining gives the highest BBD mean precision-recall area under the curve (PR-AUC) among the evaluated pretraining datasets for single-generation and multi-generation prediction. enviGraph achieves BBD single-generation PR-AUC of $0.202 \pm 0.050$ with synthetic pretraining, compared with $0.169 \pm 0.065$ for the retrospectively selected strongest USPTO condition, USPTO-Stereo (mean $\pm$ sample standard deviation across folds). Its exact BBD top-1 and top-5 accuracies with synthetic pretraining are 0.355 and 0.594. The strongest USPTO conditions give higher PR-AUC means than synthetic pretraining in most Soil and Sludge comparisons. enviGraph has higher Soil multi-generation PR-AUC means than enviFormer under each of the four main pretraining datasets, and with synthetic pretraining it also leads enviFormer on BBD single-generation PR-AUC and exact top-1 and top-5 accuracy.

High-frequency and broadband line-of-sight multiple-input multiple-output (LoS-MIMO) is a key solution for massive space-to-Earth data downlinks. However, in low Earth orbit (LEO) scenarios, it faces spatial rank deficiency from time-frequency coupling correlation and clock deviations from distributed oscillators. This paper establishes a dynamic effective channel throughput (ECT) framework incorporating waterfall characteristics of forward error correction. A hardware-friendly OFDM-based transceiver architecture with subband-adaptive truncation and dynamic timing recovery to achieve robust spatial decoupling is proposed. Through rigorous closed-form derivations, the equivalent signal-to-interference-plus-noise ratio (SINR) under joint errors is quantified, and an optimal scheduling algorithm combining truncation with variable coding and modulation is proposed to maximize ECT. Simulations confirm that the design circumvents outages caused by spatial notches with low complexity, stably enhancing the potential average throughput of a single dual-polarized beam to over 100 Gbps, providing a reliable theoretical foundation and baseband solution for next-generation high-throughput satellites.

We present J-tau, the first publicly available Japanese benchmark for evaluating tool-using LLM agents in interactive customer-support scenarios. While tau-bench is widely used to evaluate LLM agents, it is English-centric and cannot evaluate agents in other languages. Existing multilingual extensions offer no public Japanese version and rely primarily on machine translation, leaving locale-specific entities such as addresses and currencies inconsistent with the target region. To enable reliable evaluation in Japanese, we carefully localize tau-bench through machine translation, entity adaptation, and human review. During this process, we also uncover and fix two flaws in the original tau-bench: invalid tasks and a score aggregation error that inflates certain models' results. Experiments show that J-tau yields fewer evaluation errors than a naively machine-translated counterpart and that our fixes resolve both flaws.

The relationship between classical methods for solving inverse problems and neural network methods is examined. An answer is provided to the question of what inverse problems can be solved using neural networks. The answer is that neural network algorithms can solve only conditionally well-posed problems (well-posed according to Tikhonov). The reason is that the design of standard neural network algorithms violates the well-known Bakushinsky veto from the theory of ill-posed problems. It requires that any method theoretically guaranteeing a stable solution to an ill-posed problem must depend explicitly on the error level of data. The article highlights the distinction between the concept of regularization (RA) in the theory of ill-posed problems and regularization as understood in neural network training, where Bakushinsky's veto is violated. Due to this violation, any changes to the neural network architecture, regularization during training, or training dataset augmentation cannot alter the answer specified above. It is theoretically demonstrated that an ill-posed inverse problem can be solved only by using a parametric set of neural networks that explicitly depend on the known error levels (as parameters) of data. Accordingly, as an alternative to standard neural network algorithms, \neural network based regularization algorithms (NNBRAs) are proposed to solve ill-posed inverse problems. One approach to their construction involves using a regularized dataset. Such a dataset can be generated from perturbed inverse-problem data using a specific base RA explicitly depended on the error level of the data. Test solving the simplest finite-dimensional ill-posed inverse problem demonstrate that the accuracy of the NNBRA approximate solutions is comparable to that of the base RA. At the same time, errors of conventional neural network methods may be unacceptable.

Large language models (LLMs) hold broad knowledge in one set of weights, but lack interpretability and trustworthy reasoning, especially in high-stakes use cases. A neuro-symbolic model, with its interpretable architecture and System 2 reasoning capacity, is a promising alternative if built at the scale of LLMs; however, its symbolic backbone is usually constructed with human effort, so existing neuro-symbolic models are bottlenecked by the scope of the symbol component and usually domain-specific. A data-driven method could instead build a larger symbolic system by aligning and merging existing ones. We target its first, foundational step: matching units across these heterogeneous systems. Our method, SAE-IDF, encodes a symbol in its native context into a sparse code with a sparse autoencoder (SAE), re-weights each dimension by its rarity and matches across systems. On 12 tasks across 6 kinds of symbolic system, SAE-IDF outperforms the same LLM's dense baseline on all tasks with Gemma-2-9B and all but one with Llama-3.1-8B, by up to 45 points, consistently across SAE layers and configurations. SAE-IDF is also interpretable: every dimension of the sparse code has a human-readable meaning, so each match decomposes into named concepts (features). Our work paves the way to a universal symbolic backbone.

Gaze coordinates attention in human interaction through coupled eye, head, and eyelid motion. A humanoid robot must choose where to look while generating a suitable movement toward that target; fixed tracking rules alone do not model the temporal structure of human gaze. We separate these tasks with an asynchronous vision-language model that selects a scene referent, a reactive perception module that maintains fixation, and a Realistic Gaze Transformer (RGT) that generates target-conditioned motion. RGT is a Transformer-VAE trained on markerless human recordings retargeted to the robot's 11-DoF joint space: self-supervised pretraining encodes recent full-head history, and autoregressive fine-tuning exposes the model to its own predictions during rollout. At inference, a four-dimensional binocular-eye trajectory conditions coordinated neck and eye motion; an auxiliary blink predictor triggers eyelid templates during playback. On a two-second benchmark in which all controllers receive the same ground-truth target, RGT-Base obtains the lowest reported overall seven-joint MSE, and physical demonstrations show the integrated targeting and control pipeline. The evidence establishes robot-space kinematic fidelity and real-time RGT inference; perceived naturalness and open-vocabulary target-selection accuracy remain questions for controlled evaluation.

By means of $W$-representation, we investigate the Wishart-Laguerre $\beta$-ensemble and present a compact expression for Catalan polynomials. For a special Nekrasov partition function associated with the Wishart-Laguerre $\beta$-ensemble, we show that it can be expressed as a exponential form of the special Catalan polynomials. Furthermore, we make a rule to calculate the Catalan polynomials by means of Feynman diagrams and present another expression for Catalan polynomials. We also establish the connection between certain Catalan polynomials and knot theory.

Let $L$ be a linear second-order elliptic operator satisfying a nondegeneracy condition. We establish sharp $L^q$ estimates for solutions of the adjoint problem associated with $L$ and, by duality, derive sharp $L^p$ estimates for solutions of $Lu = f$. In particular, this confirms a conjecture of Carlo Pucci.

Characterized by central density depletion, bubble nuclei represent an important class of exotic nuclear structures. Collective flow in the $^{40}$Ca+$^{9}$Be projectile fragmentation reaction at 140 MeV/u is investigated using the extended quantum molecular dynamics (EQMD) model with a parameter on of the isoscalar monopole excitation strength $\varepsilon$. The projectile $^{40}Ca$ is initialized as a relaxed ten-$\alpha$ cluster configuration exhibiting a bubble-like density distribution. The results reveal that increasing $\varepsilon$ modifies the density distribution and the size of the depleted central region of bubble, producing pronounced changes in fragment yields and collective flow around $\varepsilon\approx0.5$. Below this value, heavy-fragment production plays a significant role, while the directed flow $v_1$ of protons and neutrons remains predominantly negative, indicating preferential emission toward the target side. Above $\varepsilon\approx0.5$, enhanced expansion of the central low-density region suppresses heavy residues and drives $v_1$ toward positive values. In contrast, preformed $\alpha$ clusters show different responses, reflecting their origin in the initial cluster configuration rather than solely in secondary decay. At low excitation, the cluster flow exhibits a pronounced impact-parameter dependence, whereas stronger excitation weakens cluster binding and suppresses directional anisotropy, driving the flow close to zero. The rapidity slopes of the flow coefficients also vary systematically with $\varepsilon$, tracing changes in emission anisotropy during expansion. These findings extend traditional static descriptions by probing the dynamical evolution of exotic structures in heavy-ion reactions and the contrasting collective-flow behavior of preformed clusters and single nucleons constitutes a sensitive dynamical probe of bubble-like structures.

Active flows are often confined by soft, deformable boundaries that they can push against, reshape, and in turn get reorganized by. Motivated by this interplay, we combine theory and numerical simulations to analyze the morphodynamics of an elastic vesicle enclosing a dilute suspension of extensile active particles. We model the two-dimensional vesicle as a weakly extensible elastic ring with bending rigidity, immersed in a Stokesian fluid, and describe the confined active suspension using a kinetic theory. The vesicle and the suspension are coupled hydrodynamically, through the kinematic and dynamic boundary conditions at the interface, and energetically, through a single alignment interaction that simultaneously aligns active particles with the membrane and the membrane with the particles. We show that this interaction reduces to a weak-anchoring boundary condition familiar from passive liquid crystals and derive a closed-form quiescent equilibrium state in which the nematic order in the suspension organizes into a vortex with a +1 defect core. A stability analysis of this state predicts an ordered hierarchy of activity thresholds at which successive shape modes of the vesicle are destabilized, as well as a purely passive, anchoring-driven instability of floppy vesicles. Nonlinear simulations, performed with a volume-conserving immersed boundary method, reveal that each threshold crossing gives rise to a distinct emergent state: a static spindle with boundary-pinned defects, a rotor that tank treads against its own rotation, a fluctuating spinner, and a turbulent state in which the vesicle acts as a self-stirring compartment. Beyond demonstrating how boundary deformability sculpts, and is sculpted by, confined active matter, our results open the door to study hydrodynamic interactions between such active vesicles and to use membrane fluctuations as a probe of activity of suspensions.

Automated materials discovery requires evaluating large pools of candidate configurations, with density functional theory (DFT) as a major throughput bottleneck. Pretrained machine-learned interatomic potentials (MLIPs) approximate DFT cheaply, but their errors vary by orders of magnitude across configurations, making it difficult to decide when to trust the MLIP predictions and when to fall back to DFT. We frame this decision as a risk-controlled selection problem and solve it within a conformal risk control (CRC) framework. Each configuration is mapped from atomic numbers and Cartesian coordinates to a fixed geometric descriptor, which a lightweight predictor uses to estimate an MLIP-DFT disagreement score in energy, average force, and maximum force. CRC then calibrates this score into a selection threshold for DFT labeling such that the MLIP error on the unselected configurations, which keep their MLIP labels, stays below a user-chosen fraction $\alpha$ of the expected error. Under an identical guarantee across eight dataset-MLIP pairs, our predictor routes 7% and 14% fewer configurations to DFT at $\alpha=0.3$ than committee baselines. Because committee disagreement overlooks errors the MLIPs share, our gains are largest when the DFT labels and MLIP training data differ in DFT functional and basis set, and narrow when the two agree. Without model access, the structure-only predictor also ranks errors as well as the target MLIP's latent representation and outperforms widely used descriptors such as Smooth Overlap of Atomic Positions (SOAP) and Atom-Centered Symmetry Functions (ACSF).

We compute the static quadrupole Love number $k_2$ and the dimensionless tidal deformability $\Lambda$ of cold white dwarfs in Eddington-inspired Born-Infeld (EiBI) gravity. Matter couples to the physical metric while perturbation equations are solved in an auxiliary metric, so the physical tidal field near the surface depends on derivatives of the equation of state. This dependence is regular for a surface polytropic index $1<\Gamma_{\rm env}<3/2$, but not for the $\Gamma = 5/3$ Chandrasekhar equation of state. The mechanism, an algebraic map between the metric that carries perturbations and the metric that matter feels, is common to Palatini-type theories, so the same type of surface condition should apply to tidal calculations in this wider class. We use thermodynamically consistent envelopes that exactly reach zero enthalpy. At five masses from \(0.6\) to \(1.3\,\Msun\) and five sampled Born-Infeld couplings \(\kappa\), with $|\kappa|\le5\times10^6\,\mathrm{m^2}$, the EiBI-GR differences are approximately linear in $\kappa$. The largest absolute fractional shifts are \(3.89\times 10^{-5}\) in the stellar radius $R$, \(4.09\times 10^{-5}\) in $k_2$, and \(2.36\times 10^{-4}\) in $\Lambda$. Across nine envelope models, half the spread of each quantity relative to the baseline envelope is at most \(0.0186\) in $R$ and \(0.0857\) in \(k_2\), but only \(3.02\times 10^{-5}\) in \(\Lambda\) as the changes in \(k_2\) and \(R^5\) largely cancel. A high-precision calculation shows that the envelope dependence of the EiBI-GR difference in \(\Lambda\) is small. Half the spread of this difference ranges from \(1.067\times 10^{-12}\) to \(6.208\times 10^{-11}\) over the 20 nonzero mass-coupling points.

Recent work has explored visual reasoning through image generation, where models solve visual-reasoning problems by directly generating solution images. Intuitively, iterative samplers such as flow matching are well suited to this task, since each sampling step may resemble one reasoning step. However, in this work, from a circuit complexity perspective, we show that this intuition is misleading. We prove that as long as the model outputs satisfy certain mild regularity conditions, the trajectory of the vanilla flow-matching sampler can be simulated by a constant number of sequential rounds of parallel model evaluations. Consequently, additional sampling steps cannot increase the expressive power of a constant-depth backbone model under the standard complexity-theoretic model of transformers. We then reveal that self-conditioning, which conditions the model on the previous step's estimate of the final image, plays a previously unrecognized role in removing this bottleneck. Controlled experiments on diverse visual-reasoning tasks show that flow matching with self-conditioning achieves higher accuracy and greater reasoning depth than vanilla flow matching, and its performance improves with more sampling steps. Theoretically, we prove that with self-conditioning, a constant-depth backbone can achieve expressive power equivalent to that of space-bounded Turing machines. Our study suggests a design principle for future visual reasoning research: carry more information between sampling steps, so that additional steps translate into deeper reasoning.

Differentiable reinforcement learning shows strong promise for robot policy learning, but its application to legged locomotion has remained limited to blind walking or simulated 2D tasks. Extending it to end-to-end visual obstacle avoidance is challenging because hard collision constraints and early termination upon contact corrupt first-order gradients, trapping optimizers in local minima. We present E2DRL, which addresses this with a soft differentiable obstacle-interaction formulation: penetration penalties across multiple body keypoints with velocity-conditioned stop-gradient shaping to provide the first-order training signals for 3D obstacle avoidance. To realize a deployable perception-control system, a privileged teacher trained via Soft Analytic Policy Optimization (SAPO) in a differentiable simulator is distilled into a recurrent depth-based student via DAgger, then blended with a base locomotion policy using a terrain-aware gating module for deployment. In simulation, E2DRL achieves higher sample efficiency than PPO and demonstrates superior velocity-tracking performance and 3D obstacle-avoidance capability compared with both PPO and Duality-based DCBF NMPC. We present the first sim-to-real validation of end-to-end differentiable RL for 3D obstacle avoidance on a Unitree Go2 in static and dynamic environments.

We investigate surjective isometries between the unit spheres of facially symmetric spaces and their interaction with the underlying facial structure. It is shown that such isometries preserve the essential geometric relations associated with symmetric faces, orthogonality, and geometric tripotents. In particular, a surjective sphere isometry induces a natural correspondence between the geometric tripotents of the underlying spaces and preserves the principal structural features determined by the Peirce decomposition. These results provide a geometric framework for studying linear extension problems for sphere isometries. As an application, we establish a positive solution of Tingley's problem for finite-dimensional real neutral strongly facially symmetric spaces satisfying the joint Peirce property. We also obtain an extension result for a wider class of real neutral strongly facially symmetric spaces admitting an appropriate approximation by finite-dimensional Peirce subspaces. In these settings, every surjective isometry between unit spheres extends uniquely to a surjective real-linear isometry of the ambient spaces. The results demonstrate that the facial and tripotent structures play a decisive role in the rigidity of spherical isometries.

We propose DAGFormer, a Transformer language model whose cross-layer connections are planned by a jointly trained topology predictor. Beyond the residual connections in a standard Transformer, DAGFormer adds direct connections from earlier layers to later ones, and the predictor controls the strength of every connection based on input tokens. Unlike existing cross-layer routing architectures, which set each layer's connections separately, DAGFormer plans the routing of the whole model at once. On matched OLMo-2 pairs from 75M to 1B parameters, DAGFormer lowers WikiText-2 perplexity by 17% to 25% over the vanilla Transformer, and a 1B DAGFormer trained on 5B tokens outperforms the vanilla model trained on 10B. Against four routing baselines with the same data and budget, it reaches the lowest perplexity at 75M, 300M and 1B. The gain carries over to 13 commonsense, science, reading and math evaluation tasks (2.0 to 2.5 points higher mean accuracy at 300M and 1B) and to instruction finetuning (63% to 70% AlpacaEval win rate against the vanilla Transformer, leading every routing baseline), and under structured pruning DAGFormer degrades the least in five of six settings.

Visual recognition models usually process a fixed view under a fixed computation budget, although images often benefit from different scales and amounts of spatial evidence. Active vision can adapt by observing a coarse scene and then selecting informative fixations, but the public AdaptiveNN-DeiT-S pipeline still fixes both field of view and observation count. We extend this fixed-step recognizer with a controllable observation interface that adapts both the field of view and the number of fixations to the input. Specifically, the model predicts a per-fixation scale that sets its field of view, refines each selected crop with optional scale-gated foveation and a residual high-resolution detail-token branch, and uses label-grounded stop heads for input-dependent early termination. Experiments across fine-grained benchmarks, full ImageNet-1K, and ImageNet-A/R-derived classification splits show that the interface remains competitive with fixed-step references while exposing a controllable observation budget. The same checkpoint enables threshold-controlled accuracy-compute trade-offs without per-threshold retraining. On sequential ImageNet-1K, the learned stop rule outperforms fixed-step truncation of the same checkpoint, sustaining 81.9% top-1 using 2.45 of the four focus steps and exceeding three-step truncation at lower cost. On each benchmark, a variant combining scale control and detail-token fusion achieves the highest accuracy among the tested component-added non-exit variants. Our code is available at https://github.com/GZHU-DVL/LI-Dikai.

Single-image 3D reconstruction has advanced rapidly, yet standalone garments remain challenging due to thin-shell geometry, open boundaries, and layered structures. Existing datasets and evaluation protocols provide limited support for systematic evaluation in this setting. We introduce MagicGarment, a benchmark for single-image standalone 3D garment reconstruction with 1,518 professionally constructed assets across 17 composite garment types, aligned multimodal annotations, and a unified evaluation protocol. We further establish a garment-aware reference baseline by adapting a pretrained image-to-3D backbone at two levels: Grounded Semantic Refinement (GSR) provides garment-focused visual conditioning, while Flow-Time-Route LoRA (FTR-LoRA) adapts the frozen 3D generative backbone using flow-timestep, token-routing, and expert-load information. Experiments demonstrate MagicGarment's benchmark utility and the effectiveness of garment-aware adaptation. Project page: https://1019538157.github.io/MagicGarment/.

In this paper we introduce the model of Percolative MDLA (PMDLA) on $\mathbb{Z}^2$ which engulfs the entire occupied percolation component of each particle that reaches the aggregate. We prove a phase transition of finite time explosion, as a function of the particle density $\lambda$, where the critical value equals $\lambda_c=-\log(1-p_c^{\rm site}(\mathbb{Z}^2))$. Below $\lambda_c$, the process is nonexplosive and has a linear upper bound on its radius. For $2/3<\lambda<\lambda_c$, it also has a linear lower bound on the growth rate. Above $\lambda_c$, an infinite component is swallowed in finite time a.s. At $\lambda_c$, every update is finite, but the updates accumulate in finite time, causing a blowup. Excluding criticality, the results are generalized to all higher dimensions.