Papers

While recent video generative models can synthesize high-fidelity videos, they struggle to portray plausible physical interactions and the resulting state transitions, a critical bottleneck for applications in robotics and VR/AR. To address this, we introduce a framework to generate a scalable synthetic dataset of controllable interactions. Our pipeline leverages a structured taxonomy and state-of-the-art image editing models to create explicit `start' and `end' state images, which serve as visual anchors for the interaction. To generate a seamless video utilizing these anchors, we propose State-Guided Sampling (SGS), a novel sampling technique that mitigates artifacts common in naive conditional generation. Furthermore, we develop and validate a new automated evaluation system that aligns with human judgments to ensure data quality. Experiments show that fine-tuning a base model on our dataset significantly enhances its ability to generate plausible interactions.

Haar-random unitaries provide a canonical model of generic random quantum evolution, but they typically have exponential description and circuit complexity. Although pseudorandom unitaries efficiently emulate a single Haar-random unitary, many tasks require an entire reusable family of independently random-looking operations. We introduce pseudorandom function-like unitaries (PRFU), which generate such a family indexed by public labels using only a single short key. We distinguish classical- and coherent-label access and establish security under adaptive quantum queries. For classical labels, we provide a generic construction from a post-quantum pseudorandom function and a pseudorandom unitary. For coherent labels, we use an indexed path-recording framework to analyze interference across labels and prove security of a construction based on quantum-secure function and permutation primitives. We further show that PRFUs generate pseudorandom channels and quantum combs with private memory, which are secure against adaptive interventions and concurrent sessions. Combining PRFUs with unitary gluing yields a one-key family of pseudorandom unitaries whose supported register widths can be chosen after key generation. Further applications include nonce-resolved quantum authentication, coherently masked QRAM queries, and efficient emulation of random multi-time dynamics. Our results extend quantum pseudorandomness from individual operations to efficiently generated families of random-looking quantum dynamics.

Optimizing merging coefficients for large language models can require many costly benchmark evaluations. We propose Simplex-Lattice Interpolation Merging (SLIM), which constructs a quadratic surrogate of aggregate performance on the coefficient simplex using a classical mixture design. Evaluations of individual experts and equal-weight pairs determine the surrogate with the minimum number of measurements needed to identify a general quadratic on this domain. SLIM then optimizes the surrogate without further target-metric evaluations. Experiments on two model architectures demonstrate accurate prediction of unseen multi-expert mixtures and competitive merge performance under limited evaluation budgets. Matched-budget comparisons show that structured evaluation points improve prediction fidelity over random designs, including those using regularized fitting.

Equivariant atomistic foundation models provide broadly transferable interatomic potentials trained against quantum-mechanical reference data, but their repeated execution at simulation scale remains computationally and memory intensive. We present Symmetrix-XL, an inference engine that scales pretrained MACE checkpoints without retraining, distillation, or modification of their learned weights. It combines streamed-edge execution to avoid graph-wide materialization of expanded edge intermediates, model-specialized code generation to compile checkpoint-specific operators, and tiled execution to reuse a bounded device-memory workspace. On complete LAMMPS-step benchmarks, Symmetrix-XL reduces inference time by 3.1-5.0 times relative to ML-IAP + cuEquivariance across tested A100 and RTX 5090 workloads. On a single A100 80 GB GPU, the tested MACE-OMAT-0 capacity boundary increases from 24,565 atoms to 11.24 million atoms. This substantially lowers the hardware threshold for simulations that would otherwise require spatial decomposition across many GPUs and compute nodes. The same backend weak-scales to 703 million atoms on 64 A800 GPUs at 93.8 percent efficiency. Energy, force, stress, molecular-dynamics stability, and Matbench Discovery evaluations reproduce reference MACE behavior within measured tolerances. Case studies spanning solid-state, interfacial, and reactive systems demonstrate the increased capacity in realistic workflows. These results establish post-training execution as a scaling axis complementary to model redesign, compression, and distributed scale-out, and show that the practical accuracy-cost frontier depends on both model architecture and inference execution.

We prove new large value estimates for the Riemann zeta function on the critical line. For instance, we improve the upper bound on the measure of $t\in [T, 2T]$ with $|\zeta(1/2 + it)| \geq T^{1/8}$ for the first time since Hardy-Littlewood (1923). Our results imply improved upper bounds on the $k$-th moment of zeta for every $4 < k \leq 12$. We show in particular that \begin{equation*} \int_{0}^{T} |\zeta(\tfrac{1}{2} + it)|^6 \,d t \ll_\varepsilon T^{\frac{5}{4} - \frac{1}{60} + \varepsilon}. \end{equation*} The main ingredient is a new large value estimate for exponential sums with square-root phases, and certain perturbations. Those arise in the expansion of the short second moment of zeta.

Bayesian inference increasingly uses informative but implicit priors represented only by samples, such as historical ensembles, simulator outputs, and pretrained generative models. The same computational problem appears in the test-time guidance task (generalized Bayes), where an explicit positive weight, e.g., an exponentiated reward, tilts an implicit prior. We develop a framework for source-space generalized Bayesian inference that combines inexpensive few-step prior transports with posterior stability guarantees. Specifically, we represent the prior using a one- or few-step improved MeanFlow (iMF) map and perform posterior sampling in its Gaussian source space. We establish Wasserstein error bounds between the exact and learned posteriors in terms of the joint population iMF and auxiliary-velocity loss, decomposed into training suboptimality and model-class approximation error. In the iMF source space, we adopt parallel tempering with preconditioned Crank-Nicolson updates and introduce a hybrid variant that incorporates split Hamiltonian Monte Carlo to improve sampling efficiency. Synthetic experiments show that the proposed framework can approximate posterior distributions accurately and efficiently, while CLIP-guided ImageNet experiments demonstrate its ability to steer a pretrained iMF image prior toward text-specified preferences.

We introduce GeoFWI3D, a large-scale open-source benchmark dataset of geologically plausible 3D subsurface models designed to accelerate deep learning (DL) assisted seismic imaging and full waveform inversion (FWI). FWI is a physics-driven, wave-equation-based optimization technique used in seismic imaging to estimate the subsurface properties. However, traditional FWI is highly non-linear, non-unique, and ill-posed. The iterative process often converges to local minima when the starting model is insufficiently accurate. Most importantly, 3D FWI is prohibitively expensive. With the advent of DL, a direct mapping between the shot gathers and subsurface properties can be established by training a network on realistic models. The primary bottleneck for such approaches is the lack of large-scale, realistic training datasets. GeoFWI3D addresses this gap with 10,000 geologically diverse velocity models at 96x96x96 resolution, spanning four structural complexity classes: pure stratigraphy, faulted networks, salt diapirism, and complex coupled fault-salt systems. Each model is accompanied by co-registered multi-modal labels including compressional velocity (Vp), zero-offset seismic reflectivity, relative geologic time (RGT), and semantic fault/salt masks. To facilitate systematic evaluation, we introduce benchmark tasks covering fault detection, joint salt body segmentation and chronostratigraphy prediction, FWI, wavefield and traveltime surrogates using neural operators, and generative modeling with a 3D diffusion model. We present baseline results for each task to establish reference performance metrics for future users. The dataset is publicly available under Creative Commons Attribution 4.0 International.

We present the results of a spectroscopic and multi-band photometric study of the magnetic cataclysmic variable 1RXS J174320.1-042953. Multi-band photometry confirms a periodic modulation with an orbital period of $P_{\rm orb}=0.0864 \pm 0.0001$ days. Using the estimated secondary mass of $\sim0.15\,\mathrm{M_\odot}$ and, the X-ray-based white dwarf mass of $\sim0.75\,\mathrm{M_\odot}$, we obtain values of $q\sim0.2$ and a system inclination of $47^\circ\pm10^\circ$. The optical spectra of 1RXS J174320.1-042953 are dominated by strong, complex, asymmetric, and highly variable single-peaked emission lines of the Balmer series, \ion{He}{i}, and \ion{He}{ii}, which are characteristic of polar-type magnetic cataclysmic variables. Doppler tomography reveals no evidence of a Keplerian accretion disc, supporting the classification of the system as a polar. The H$\beta$ and \ion{He}{ii} emission lines can be decomposed into at least two distinct components. The low-velocity components have semi-amplitudes of approximately 170 km s$^{-1}$ (H$\beta$) and 147 km s$^{-1}$ (\ion{He}{ii}), while the high-velocity components reach about 317 km s$^{-1}$ and 460 km s$^{-1}$, respectively. The low-velocity component is likely associated with the irradiated side of the secondary star facing the white dwarf and/or the vicinity of the L$_1$ point, whereas the high-velocity component is related to the accretion-stream structure. A redshifted absorption component in the emission-line wings is detected at orbital phases near $\phi \approx 0.0$, reaching velocities up to $\sim1400$ km s$^{-1}$ and likely produced by accretion-stream material crossing the line of sight.

Behavioral models of nonlinear microwave devices – the Cardiff model, X-parameters, the higher-order describing functions of the mechanical-systems literature – all share a functional form. This tutorial derives that form from time invariance alone. It develops the consequences of that derivation using only the harmonic-balance description of a single-tone periodic steady state, in which each port waveform is a finite set of harmonic phasors, and uses them to give the Cardiff model's three exponents physical interpretations. In the monomial rewriting the magnitude exponent $m$ is the order of the device's load-side nonlinearity; the phase exponent $n$ is set by the drive-side harmonic; and the conjugate index $r$ obeys $r_{\max}=\lfloor K/2\rfloor$, where $K$ is the degree of the load-side nonlinearity. The familiar restriction $r\le1$ is therefore a statement about the device, exact whenever $K\le3$. The final sections show that a tailored A-pull measurement displays this decomposition directly: each spectral cluster's half-width is the order of the nonlinearity that produced it. The tutorial is pedagogical and is meant as a gentler introduction to the results treated in the companion paper "Time invariance, circle symmetry, and the completeness of the Cardiff behavioral model", which states and proves the theorems in full.

The ASF is a major source of positioning error in eLoran system and is affected by terrain characteristics along the signal propagation path. This study analyzes the correlations between spatial ASF and path-based terrain features using NASA SRTM 30m and NGII 90m DEMs. Path mean elevation and path mean slope were extracted along the propagation paths from the Gwangju eLoran transmitter to measurement locations in the Incheon and Pyeongtaek port areas. The results show that both terrain features are strongly correlated with spatial ASF, with the higher-resolution NASA SRTM DEM generally exhibiting stronger correlations, particularly for path mean slope. These findings demonstrate the importance of DEM spatial resolution in terrain-based spatial ASF analysis and provide a basis for selecting terrain data for future machine-learning-based ASF prediction

We prove that the weak Pareto solution set of a multi-objective optimization problem with convex polynomial objectives and constraints is connected, assuming neither boundedness of the constraint set nor a constraint qualification. Since this set is semi-algebraic, its connectedness implies that it is path-connected whenever it is nonempty. In particular, convex quadratic multi-objective optimization problems have connected weak Pareto solution sets; this answers the weak Pareto part of an open question posed by N.~D.~Yen in Recent Developments in Vector Optimization (Springer, 2012).

Machine learning models often suffer performance degradation under subpopulation shift, particularly when spurious correlations cause models to rely on shortcut features that fail to generalize across subgroups. A recent line of work mitigates this issue by using loss-based signals to identify informative samples, but these signals can become severely distorted under label noise: mislabeled samples may also incur large losses and contaminate subsequent reweighting or retraining. Despite its practical importance, this intersection remains largely underexplored. We propose POTER, a reweighting framework based on optimal transport that derives sample importance from the transport geometry between the training distribution and a reference distribution constructed from limited validation group annotations. By measuring alignment at the individual-sample level rather than relying on loss, POTER downweights mislabeled or strongly bias-aligned samples while assigning higher importance to samples better aligned with the reference distribution. In addition, POTER requires only a single ERM training stage, moving beyond the retraining paradigm common in recent work. Across standard benchmarks and noisy-label settings, POTER achieves state-of-the-art worst-group accuracy, including cases where label corruption is concentrated within minority subgroups.

Recent advances in Large Language Models (LLMs) and Retrieval-Augmented Generation (RAG) have significantly democratized access to legal information. Nevertheless, most existing legal assistants remain confined to multi-turn conversational QA, failing to support complex legal tasks that require systematic evidence retrieval, multi-step reasoning, and report-level synthesis. In this paper, we present LawCompass, an evidence-grounded legal assistant that navigates the transition from standard Legal QA to multi-agent deep research. LawCompass provides three task-oriented functions: Legal QA, which delivers precise, evidence-backed answers to legal questions; Professional Retrieval, which enables structured exploration of statutes and judicial cases via query rewriting; and Deep Research, which employs a multi-agent workflow to decompose complex legal tasks and synthesize comprehensive research reports. Crucially, LawCompass maintains explicit citation links across all modules, empowering users to directly verify system outputs against original legal sources. Evaluation results demonstrate that LawCompass provides a practical and scalable paradigm for transforming conversational AI into trustworthy and evidence-grounded legal research assistance.

Tackling complex real-world tasks can exceed the capabilities of a single large language model (LLM), motivating the use of multi-agent workflows that coordinate specialized agents to work together on these tasks. Recent methods train LLMs to construct better workflows from execution outcomes, but they optimize only the workflow generator, while the other agents that build or execute each workflow remain fixed even though every outcome depends on all of them. However, extending training beyond the generator is challenging: the agents are coupled, and a workflow's outcome is a single sparse score that cannot tell which agent causes a failure. We propose FloWright, which leverages the workflow as a harness to optimize workflows. By introducing a hierarchical, structure-aware reward paradigm, FloWright enables one role to self-evolve and two or more roles to co-evolve, with no additional models, labels, or executions. Considering the limitation that workflows are commonly trained and evaluated on data that a single agent can already handle, we further propose DataWright, an adaptive data hardening approach that converts existing datasets into workflow-level tasks with increased difficulty. Across document, slide, chart, code, math, and finance tasks, small open models trained with FloWright achieve improved performance by up to $+7.41\%$, with co-evolving ($+5.03\%$) more roles gaining more than optimizing one of them alone ($+2.83\%$). Our project page: https://xhguo7.github.io/FloWright/.

Device-independent conference key agreement (DI-CKA) lets a group of parties establish a shared secret key from untrusted quantum devices, with security certified by non-locality. Existing DI-CKA protocols are each built around a single Bell inequality, typically a multiparty variant of the CHSH game. DI-QKD protocols, in contrast, have been built from a much richer landscape of non-local games, and it has remained unclear how to carry this landscape over to the conference setting. We introduce $\textit{Parity-$G$ games}$, which extend any two-player game $G$ to $N$ players, for every $N$, provided $G$ has an optimal strategy in which one player measures Pauli observables. The extension preserves the quantum and classical values of $G$, and the security of the resulting $N$-party protocol follows from an analysis of the two-player game alone. Our framework recovers the Parity-CHSH game of Ribeiro, Murta and Wehner (Phys. Rev. A, 2018) as a special case. Applied to the Mermin–Peres Magic Square Game, it yields a new $N$-player pseudo-telepathy game, the $\textit{Parity Magic Square Game}$, which ideal devices win in every round. We use it to construct the $\textit{first}$ DI-CKA protocol based on a pseudo-telepathy game. We prove the protocol secure against coherent attacks. It produces up to two key bits per round, and at low noise its key rate exceeds that of the DI-CKA protocol based on the Parity-CHSH game.

The $T$-count is a dominant cost of fault-tolerant Clifford$+T$ computation. We prove that every Hadamard-free Clifford$+T$ circuit with clean ancillas for a layer of $m$ disjoint CCZ gates, the diagonal core of a parallel Toffoli layer, uses at least $6m+1$ $T$ gates. Campbell and Howard gave a circuit with $6m+1$, and to our knowledge this is the first proof that it is optimal for general $m$. The proof combines Campbell and Howard's self-orthogonality of the columns of the parity matrix with a rank (flattening) bound, sharpened by one. More generally, the argument gives a lower bound on the $T$-count of every diagonal gate in the third level of the Clifford hierarchy, the isotropy floor, computable by Gaussian elimination. It is never below the stabilizer nullity $\nu$ and reaches $2\nu+1$ on non-Clifford gates whose phase is purely cubic. No bound that also holds for circuits with measurement and feedforward can do this, since such circuits implement CCZ ($\nu=3$) with four $T$ gates and the layer with $4m$, where the floor gives seven and $6m+1$. The floor also recovers Campbell and Howard's count $4m+3$ for a fan-out of Toffolis. On the output of PyZX's TODD-based optimizer, relative to its Hadamard placement, it certifies $193$ of $311$ phase-polynomial blocks $T$-optimal once repeated parities are merged (at the best of five optimizer seeds), against $113$ for nullity, including $37$ of the $72$ blocks too large for exhaustive search.

Coding agents can return plausible patches that omit required behavior. These failures are hard to review because long traces and confident summaries often hide what was missed. We ask when a nominally weaker reviewer can reliably decide whether a patch solves its issue. We study 411 execution-labeled traces from three agents and 101 controlled cases. On 154 GPT-5.4 traces, structured but unchecked evidence raises both defect catch and over-rejection. We then provide official execution evidence as an upper-bound diagnostic. After choosing and freezing one of two formats per reviewer, five of six reviewers improve both rates on 122 held-out traces; two classify every trace correctly. Reviewer size is not a consistent predictor of quality. Because official tests are unavailable in deployment, we also evaluate a frozen cascade with patch-caused static errors and generated tests that first fail on the unpatched repository. On 121 scored held-out GPT-5.4 traces and 59 Gemini traces, its coverage is 0.89 and 0.86, risk is 0.33 and 0.26, catch is 0.76 and 0.80, and over-rejection is 0.66 and 0.67. Most false rejections occur when unresolved cases reach the reviewer. Official execution evidence shows the potential of weak review when decisive checks are available. Producing equally reliable checks without official tests remains the main bottleneck.

Memory-attention-based Video Instance Segmentation (VIS) methods have demonstrated strong zero-shot tracking capability, yet their substantial memory requirements confine them to short video clips and their single-prompt inference design makes multi-category open-vocabulary tracking computationally prohibitive. This work introduces two contributions toward fully automated tracking annotation of arbitrary video. The Generalized Presence Token (GPT) reformulates SAM3's inference pipeline to process N text prompts simultaneously via virtual prompt batching, reducing image encoding cost from O(N) to O(1) with no modifications to any learned component. The Annotation and Segmentation Handler (ASH) extends any memory-attention VIS tracker to sequences of arbitrary length through overlapping temporal chunks with IoU-based inter-chunk identity matching, requiring no dataset-specific training. Instantiated on SAM3, the resulting pipeline – SAM3-ASH – achieves state-of-the-art HOTA on MOTS20 under fully zero-shot conditions and remains competitive with trained specialists across seven additional benchmarks, while peak GPU memory consumption stays below 25 GB, establishing a practical baseline for scalable, training-free automated video annotation.

Low Earth orbit (LEO) satellite Doppler measurements extracted from passive software-defined radio (SDR) spectrum data require temporal alignment with predicted satellite trajectories for reliable satellite association. In our previous framework, Doppler slope information was used during both segment generation and clock offset estimation, while the clock offset was estimated at a single coarse receiver position. This study proposes a satellite matching framework based on continuity-based Doppler segmentation and multi-position clock offset estimation to reduce these dependencies. Doppler segments are first generated using only temporal and frequency continuity, thereby separating segment generation from the slope criterion used for clock offset estimation. The clock offset is then estimated using five coarse receiver positions, consisting of one city-level nominal position and four surrounding positions, through a two-stage procedure based on slope compatibility and a combined Doppler cost. The aligned segments are subsequently associated with candidate satellites using a matching score based on the same Doppler slope and bias-removed RMSE metrics, and the resulting associations are evaluated through receiver localization. Experiments using six hours of passive Starlink/OneWeb monitoring data show that continuity-based segmentation alone did not improve receiver localization compared with slope-based segmentation under single-position clock offset estimation. However, applying the proposed multi-position clock offset estimation to the same continuity-based Doppler segments reduced the localization error from 32.99 km to 1.84 km, achieving the lowest error among the evaluated methods. The results for the evaluated dataset show improved spatial consistency of satellite associations when multi-position clock offset estimation is applied to continuity-based Doppler segments.

Assessing learning in large classrooms presents a significant challenge for individual instructors, who may have limited capacity to evaluate the understanding, participation, and assessment behaviour of every student. Peer assessments have been a way of distributing this responsibility among learners, allowing them to evaluate and provide feedback to one another while reducing dependence on instructor-led assessments. Building on this approach, we implemented a peer validation model within a large, multi-institutional internship programme in which students who demonstrated sufficient understanding were authorised to assess and validate their peers through short oral discussions. The assessment process began with the instructor validating a small group of students, who were then authorised to validate their peers, allowing the process to gradually expand across the cohort and operate at scale. This study examines how participants experienced the model and the extent to which assessment integrity was maintained, using an end-of-programme survey of 238 consenting respondents. Most participants regarded the activity as worthwhile, with 79.8% reporting that they solved problems they could not previously solve. However, 29.0% acknowledged at least one instance of reduced effort, a lowered validation standard, or reciprocal validation, while 88.7% believed that at least a little validation had occurred without proper examination. When asked how the process could be strengthened, participants selected post-validation discussion of solutions approximately twice as often as closer auditing or mentor-led validation. These findings provide descriptive evidence of both the potential and the integrity challenges of using peer validation as a scalable assessment approach in large learning environments.

Robotic world models predict action-conditioned future scenes, providing a foundation for understanding action outcomes. However, turning alternative predictions into useful learning signals remains challenging: similar candidates limit informative quality comparisons, while diverging trajectories require persistent maintenance of their individual histories. We introduce FutureWorlds, a framework that unifies candidate construction, history maintenance, and learning from relative quality. Built on a multimodal discrete autoregressive model, FutureWorlds uses diverse beam search during reinforcement learning to construct candidate futures that balance confidence and diversity. Candidate-specific bounded memory preserves scene states and ensures that generation and policy scoring use matching histories. We further propose MemSPO (Memory-Conditioned Search-Guided Policy Optimization), which converts video trajectory rewards into group-relative advantages to optimize the world model. On RT-1, BridgeV2, and RoboCasa, FutureWorlds reduces LPIPS for 32-frame predictions by 14.78%, 20.84%, and 9.12%, respectively, relative to the strongest baseline on each dataset. Under fixed evaluation configurations, only 200 MemSPO updates further improve generation quality and support continued prediction beyond the training horizon. Memory ablations, decoding sensitivity analysis, and optical-flow evaluation show that these gains extend beyond visual quality to more accurate motion prediction and more consistent object states. Project page and code: https://github.com/Alexander-wu/FutureWorlds.

The use of signals from Earth-orbiting Global Navigation Satellite System (GNSS) constellations for positioning, navigation, and timing on the lunar surface was recently demonstrated by the Lunar GNSS Receiver Experiment (LuGRE) aboard Blue Ghost Mission 1. For receiver clock synchronization using GNSS, Kalman filter performance depends on the satellite-specific measurement uncertainty used to construct the measurement noise covariance. Terrestrial measurement weighting commonly relies on receiver elevation angle, but the elevation angles in the analyzed LuGRE data are confined to 24.2-31.5 degrees, providing limited discrimination of measurement quality. In contrast, satellite nadir angles span 12.1-66.6 degrees, and the measurement variability increases substantially at small nadir angles. Based on this observation, we propose a measurement weighting model that combines satellite nadir angle with C/N0 for lunar surface receiver clock synchronization. The proposed model is evaluated against three weighting methods adopted from the literature using seven LuGRE operation periods (OPs). It achieves the lowest receiver clock bias root mean square error (RMSE) in six of the seven OPs and an overall RMSE of 5.78 m, which is 19.5% lower than that of the best comparison model. The proposed model also provides the lowest clock bias prediction RMSE under simulated measurement outages.

Large language models (LLMs) increasingly construct multi-agent workflows that decompose a complex task and assign specialist agents from a pool. However, building such a workflow well remains challenging: how finely to divide the task, which agent to trust with each subtask, and when to create a new specialist are all critical decisions a workflow constructor needs to settle up front. Thus, whether each subtask succeeds remains unknown until the workflow runs. Yet, improving a workflow is costly. Locating a fault usually requires a reference answer, a graded outcome, or a trained assessor, and the fix is applied to the whole workflow through re-execution, re-search, or retraining. We propose InFlowOp, which prices every decision in one label-free cost that weighs how well an agent's competence meets what a subtask demands against how much that agent takes to run. Before execution, InFlowOp bidirectionally determines the granularity of task decomposition and agent assignment following from the cost rather than from a fixed template. During execution, InFlowOp corrects a fault with the cheapest move via the same cost that serves the workflow both as it is built and as it runs. Facing the workflow-level evaluation challenge, we introduce Braid, a benchmark whose tasks require multi-agent coordination beyond single-agent capability. Across various domains and backbones, InFlowOp outperforms single agent baselines by up to $+11.97\%$, achieving $+9.64\%$ with in-flow optimization. Our project page: https://xhguo7.github.io/InFlowOp/.

Diffusion language models (DLMs) offer a promising alternative to autoregressive (AR) language generation. Recent advances in continuous DLMs, which apply latent diffusion to continuous text embeddings, raise a practical question: which embedding makes the best latent space, i.e., the most diffusible? To answer this, we search through different embeddings and find that scaling the embedding model to stronger ones within the same family (T5 to T5Gemma-1 to T5Gemma-2) greatly improves generative performance. But the raw T5Gemma-2 embeddings are still not optimal. They are so discriminative that even the embeddings of plausible alternative words are separated, which makes the generation vulnerable to imperfect sampling. Consequently, continuous diffusion often fails to reach any of them and ends up at an invalid embedding instead. To address this, we distill T5Gemma-2 into a student encoder that learns the teacher's decoded probabilities as soft labels. Learning from such soft labels makes the student pull the alternative embeddings closer while maintaining the encoding-decoding mechanism. The distilled embeddings form a more connected and diffusible latent space, improving over the vanilla T5Gemma-2 embeddings. As a result, our medium-sized DLM achieves Gen. PPL 17.8 (against real-text PPL 15.4) at real-text entropy on OpenWebText, outperforming GPT-2-M on Gen. PPL.

Photoacoustic tomography (PAT) is a hybrid imaging modality that combines high optical contrast with high ultrasonic resolution for biomedical imaging applications. In this work, we investigate the inverse problem of recovering the initial pressure distribution from boundary measurements in the presence of nonlinear acoustic propagation and viscous attenuation effects. To model these phenomena more accurately, we consider a nonlinear damped viscoelastic wave equation incorporating spatially varying sound speed, temporal attenuation, and nonlinear propagation mechanisms. We first establish the well-posedness of the corresponding forward problem using a Galerkin approximation combined with energy estimates and a fixed-point argument. For the inverse problem, we derive existence, uniqueness, and local uniqueness results under suitable assumptions through a harmonic extension reduction, spectral Laplace transform techniques, and observability estimates. To numerically reconstruct the initial pressure field, we develop a hybrid reconstruction framework that combines a convolutional neural network (CNN) with a gradient-free optimization strategy based on the sequential quadratic Hamiltonian (SQH) method derived from Pontryagin's maximum principle. The CNN is used to generate an informative initial guess, while the SQH framework enforces the governing PDE dynamics during the reconstruction process. Numerical experiments demonstrate that the proposed hybrid strategy significantly improves reconstruction quality, contrast, and robustness compared to standalone time-reversal and CNN-based approaches.

Voting over multiple LLM responses is a common primitive in test-time scaling and ensemble inference. Collecting more responses can expand the candidate pool and increase the chance that a correct answer is discovered. Under a fixed call budget, a discovered answer still needs to accumulate enough support within the remaining calls to become the final plurality winner, creating a discovery-to-decision gap. In this work, we characterize this gap through the realized vote state and remaining call budget. We derive a sharp recoverability threshold and show that, as sampling proceeds, the observed candidate set can only expand while the set of reachable endpoint winners can only contract, inducing a candidate-level conversion window. Under a specified iid response law, the same state yields exact finite-horizon endpoint probabilities. We further show that merging wrong-answer identities preserves single-call correctness and cannot improve plurality accuracy, and that the effect of redistributing wrong-answer probability depends on the realized vote state. Singleton reachability yields a gold-free exact locking certificate. For a known answer universe, its first trigger is the earliest prefix at which all admissible continuations yield the same fixed-budget output. Empirically, most discovered-but-unselected correct answers lose reachability only after discovery. In a controlled Word16 study, input permutation improves raw-plurality accuracy by 21.1 points with essentially unchanged single-call correctness. Exact locking saves 28-30% of calls at a 16-call budget while preserving every fixed-budget output.

Feed-forward 3D vision models such as VGGT have achieved remarkable progress, unifying camera estimation and dense scene reconstruction in a single pass. However, their quadratic global attention makes long image sequences expensive, while existing sparse methods may favor highly attended yet value-redundant regions. To address these limitations, we introduce VASC, a training-free sparse attention method combining value-aware block selection and execution-aware cross-layer memory. Our value-aware block selection integrates pooled query–key relevance with neighboring value contrast, reducing redundancy while preserving query-relevant and distinctive content. Cross-layer memory tracks unserved demand across layers and updates this state according to actual execution, enabling previously underserved blocks to compete under a fixed computation budget. Experiments on 7Scenes and NeuralRGB-D with VGGT and $\pi^3$ demonstrate improved pose estimation and reconstruction quality compared with FasterVGGT, together with up to $2.29\times$ faster inference than dense VGGT. Code is available at https://github.com/kosakayamahoo-design/VASC.

Video-to-speech synthesis aims to generate natural-sounding speech from silent talking-face videos while ensuring phonetic accuracy. A fundamental challenge in this task is the inherent one-to-many mapping problem, where visual dynamics often lack sufficient information to uniquely determine the corresponding utterance. To address this, we propose Watch Your Speech (WYS), a video-to-speech synthesis framework that incorporates textual conditioning as an explicit linguistic cue to mitigate visual ambiguity. Our framework features an attention-based embedding fusion module that synergistically integrates textual context with video sequences, coupled with a conditional flow matching objective for high-fidelity speech generation. Extensive experiments on the LRS2 and LRS3 datasets demonstrate that WYS achieves superior performance, establishing new state-of-the-art results in audio-visual synchronization (LSE-C/D) while maintaining highly competitive textual accuracy (WER). Subjective evaluations further confirm that our model generates speech with near-human naturalness, validating the effectiveness of textual conditioning in content-controlled video-to-speech synthesis. Project page: https://github.com/gunwoo5034/Watch-your-Speech

We investigate collective modes in charge-neutral altermagnetic superconductors with $d$-wave spin-split bands, focusing on the impact of emergent Bogoliubov Fermi surfaces. In the absence of a magnetic field, the BCS state supports collective modes that inherit the underlying $d$-wave symmetry, resulting in pronounced momentum-space anisotropy. The broken time-reversal symmetry further couples these modes to spin-density fluctuations, allowing their signatures to emerge in the spin dynamical structure factor. Upon applying a magnetic field, Bogoliubov Fermi surfaces emerge and qualitatively reshape the collective-mode dynamics through gapless particle-hole excitations and strongly anisotropic Landau damping. The damping is strongest along the momentum-space diagonal, where it can suppress the low-energy phonon mode at small momenta. Remarkably, the suppressed phonon is accompanied by the emergence of an in-gap structure in the Higgs mode and an associated pronounced enhancement of the spin response. These results establish Bogoliubov Fermi surfaces as a mechanism for generating highly anisotropic collective-mode dynamics and identify spin response as a sensitive probe of in-gap Higgs excitations in altermagnetic superconductors.

Cooperative perception, in which connected vehicles and roadside infrastructure share sensor information, is a candidate enabler of automated mobility, and benchmark accuracy is the evidence cited when roadside deployment is considered. This paper audits that evidence base across one simulated and two real-world vehicle-to-everything (V2X) benchmarks. In simulation, two widely studied robustness axes leave almost no recoverable headroom: an infrastructure-anchored pose correction returns about one accuracy point at every error level, and corrupting a partner costs 0.7 points. On real data, measurement choices govern the conclusion. An apparent seventeen-fold advantage of infrastructure in partner-poor frames falls below four-fold once the split is broadened, and an ad-hoc class definition measures a far smaller benefit than the official protocol reports. Cooperation is worth 7 to 15 accuracy points, yet 27% of frames on one split offer no partner and almost none do on another, so every regime-conditioned claim must name its split.