Papers

We present a geometry-refinement workflow that adapts to the cost of an energy–gradient evaluation and the reliability of molecular topology. An initial Cartesian calculation and continuous bonding diagnostics determine whether to begin in projected Cartesian coordinates or in chemically adapted internal coordinates. Chemical perception generates local redundant measurements; sparse rank and conditioning tests combine them into nonredundant, symmetry-adapted coordinates; and a common optimizer uses the selected representation with resident or external calculators. For inexpensive gradients, the projected Cartesian route limits coordinate overhead; a single monitored internal-coordinate step can then provide chemical interpretability. When gradients are costly and topology is certified, symmetry-adapted internal coordinates reduce expensive energy–gradient calls. A quadraticized map improves localized anharmonic cases, while mixed Cartesian/internal charts cover uncertain or complex topologies. Local sparse operators and resident force-field and tight-binding calculators extend both routes to large systems. The same Cartesian calculator contract supports multilevel energy and gradient composition. The measured coordinate pipeline and recurrent resident tight-binding update show near-linear scaling over the tested size ranges; this statement applies to those qualified components, not to arbitrary electronic-structure calculations or cold starts. For certified, directly separable internal charts, the optimizer can use chemically designed Morse or Gaussian maps to quadraticize the local step problem. The quadraticized branch uses a short geometry-local history, while delocalized charts remain unchanged. This reduces iterations in difficult minima without changing the electronic potential or coordinate chart.

A higher evaluation score does not always mean a better language model system. When optimization exploits an evaluator's mistakes, measured progress can conceal unchanged or deteriorating task performance. This failure can arise through parameter updates, selection among generated outputs, or revisions to persistent prompts. We develop a comparative framework for reward hacking across these three optimization substrates: weights, selection, and text. Building on the Proxy Compression Hypothesis and research on inference-time and in-context reward hacking, we examine how reachable behavior, optimization budgets, and persistent adaptation shape exposure to proxy error. We formalize a distance-dependent upper bound on evaluator disagreement and a capacity ordering for nested policy classes, then show why distance alone cannot establish a universal ranking of vulnerability. An exact finite-output illustration demonstrates how the location of a scoring defect changes the behavior favored by each method. We also map representative defenses across substrates, identifying which mechanisms transfer directly and which offer only functional analogies. Persistent prompts receive particular attention: their contents are inspectable, but the behavior induced by a small textual change may be difficult to anticipate. The formal analysis, numerical illustration, and published evidence together provide a basis for comparing optimization methods and identifying the conditions under which their defenses transfer. The resulting framework connects optimization choices to verification requirements: reliable improvement depends on controlling accessible failure modes and preserving evidence of task quality independent of the score being optimized.

Classical coherent binary phase-shift keying (BPSK) reception over additive white Gaussian noise (AWGN) channels is analytically well understood, and the optimum hard-decision detector is known. Therefore, the aim of using neural receivers in this study is not to replace the classical AWGN-BPSK detector. Instead, the AWGN-BPSK setting is deliberately selected as a theoretically verifiable benchmark for analyzing how compactly neural receiver architectures can represent a known decision behavior. This paper compares multi-layer perceptron (MLP) and Kolmogorov–Arnold Network (KAN) receivers using a multi-SNR Nyquist-rate BPSK dataset. The models are evaluated using bit error rate (BER), mean squared error (MSE), and trainable parameter count. Beyond reporting accuracy alone, this work emphasizes a common-loss parameter-efficiency perspective: when two receivers reach a comparable practical BER or loss region, the receiver with fewer parameters is more attractive for real-time deployment. The results show that both MLP and KAN receivers reproduce the expected AWGN-BPSK detection trend, while a compact KAN configuration reaches a comparable operating region with substantially fewer trainable parameters. In particular, the KAN receiver with 3281 parameters achieves a test BER of 2.45 x 10^-4, while the MLP baseline with 8513 parameters achieves a test BER of 2.50 x 10^-4. This corresponds to approximately 61.5% fewer trainable parameters at a comparable BER operating point. This reduction is important for real-time neural receivers because it affects memory footprint, parameter access, inference latency, energy consumption, and hardware feasibility on embedded, software-defined radio, FPGA, ASIC, and edge communication platforms.

We study correlation functions of giant and dual giant gravitons in $\mathcal N=4$ super Yang–Mills theory. Although the two objects correspond to rather different D3 brane configurations in $AdS_5\times S^5$, their gauge theory operators are Schur polynomials associated with transposed Young diagrams. We show that this transposition is accompanied by a formal rank reflection $N\to -N$ at the level of correlation functions. For both $U(N)$ and $SU(N)$, the relation is exact for free correlators at finite $N$, and, using Lagrangian insertion, extends to separated point weak coupling integrands at every loop order. More generally, it applies to any number of Schur polynomial external operators, provided all Young diagrams are transposed. We illustrate the relation in several examples and also discuss evidence for its extension beyond perturbative integrands.

Many vision applications ask several independent, forced-choice questions about the same image. Visual Jev encodes the image and public context once, executes isolated question suffixes as a batch, and reads candidate probabilities from the backbone's language-model head. Across four benchmarks, answer-supervised post-training raises equal-weight macro accuracy from 70.6% to 76.1%, with the gain concentrated on the two task families represented in training. At N=32 questions per image, shared batched execution is 8.9x faster in warm amortized time than independent serial execution and remains 3.4x faster than an already-batched baseline that recomputes the prefix, at the cost of higher peak memory. A matched typed-head control offers no consistent accuracy advantage over the language-model-head readout. The supported design is therefore simple: adapt the backbone for quality, retain the existing readout, and share execution for efficiency.

Measurement and classical randomness usually degrade coherence; controlling where and when a subsystem is monitored can instead make them a tunable resource. We introduce a photonic time-multiplexed quantum walk whose fast reconfigurability enables precise position measurements, allowing us to study this interplay in a hybrid quantum-classical pursuit problem: A quantum walker (lamb) evolves on a line while a classical pursuer (lion) performs a lazy random walk. Each lion trajectory drives a distinct coherent, norm-reducing evolution of the lamb, with capture implemented by measurements at the lion's positions; averaging over the recorded trajectories then synthesizes a decoherent process. Counterintuitively, increasing the lion's mobility does not monotonically suppress survival: over a broad range of hopping probabilities, the lamb survives better than against an immobile predator. Directly observing this local enhancement, we demonstrate a manifestly quantum feature aided rather than hindered by classical stochastic control.

We prove that the variety generated by the max-plus algebra on the nonnegative integers has a unique limit subvariety, namely the variety generated by the six-element semiring constructed by Shao, Ren and Gao. More generally, we characterize hereditary finite basedness in a finitely defined variety of commutative additively idempotent semirings containing the max-plus variety. Its hereditarily finitely based subvarieties form a finite lattice and admit equational bases involving at most ten variables. We also prove that every proper subvariety of the max-plus variety is locally finite. A finite basis theorem for finite semirings with a multiplicative identity establishes finite basedness of all finite truncations. Finally, we determine the finite basis properties of two families of subvarieties defined by power identities and of subvarieties relatively defined by arbitrary families of identities making individual additive subterms greatest elements.

We consider whether the kink instability can explain particle acceleration in hard state 'compact' X-ray binary jets. We apply a 1D model to assess the expected growth rate of the instability for a variety of jet models, then scale results from simulations to estimate a dissipation rate for the magnetic energy in the jet. We then consider the effects of radiative cooling on the system. We find that hard state 'compact' X-ray binary jets are likely to be unstable to the kink instability within their acceleration zones across a plausible parameter space. We further find that depending on the parameters of the model, the jets may radiate over a variety of length scales ranging from length scales attributable to a 'failed' jet scenario, to those in keeping with resolved radio observations of 'compact' jets. We discuss the expected evolution of this model through the outburst cycle.

Metric reasoning is a critical and challenging task for Vision Language Models (VLMs), playing a pivotal role in embodied AI tasks such as robotic manipulation and autonomous navigation. However, current spatial reasoning remains bottlenecked by rigid pixel-level supervision; such localized optimization often compromises general multimodal intelligence, triggering performance degradation or catastrophic forgetting of broad reasoning capabilities. To address these limitations, we introduce Metric-Bench, a focused benchmark designed to guide metric-spatial reasoning using contextual information. By incorporating in-image reference objects with known physical dimensions, Metric-Bench guides models to implicitly learn the 2D-to-3D mapping without camera intrinsics. We further present MetricReasoner, a task-adapted reinforcement fine-tuning recipe for reference-grounded metric reasoning, using structured prompts and verifiable numerical rewards. Extensive experiments on Metric-Bench demonstrate that our approach significantly enhances spatial metric understanding, outperforming existing and even larger proprietary models by 43.1%, while improving downstream embodied performance over a spatial-specialized counterpart by 30.4% on RoboSpatial overall accuracy and 9.3% on ERQA, and additionally delivering consistent gains on general benchmarks (15.9% on V$\star$Bench, 88.9% on BLINK), indicating that the proposed adaptation does not necessarily compromise general VLM capabilities.

We obtain nonperturbative asymptotic expansions of the electromagnetic form factors of the pi and rho mesons at large spacelike momentum transfer, Q^2->\infty. The calculation is based on a double integral representation in the invariant masses of the composite system. The model parameters are kept at the values fixed in the authors' earlier description of meson electroweak properties. We show that, in the limit of vanishing constituent-quark mass, with the constituent quarks taken to be pointlike, the asymptotic ratio of the longitudinally polarized rho-meson to pion form factors in the helicity basis is consistent with the prediction of perturbative QCD (pQCD) within the adopted parameter ranges.

Flow-matching schedules affect both sampling dynamics and the variance of the regression target. For centered commuting Gaussians, we show that a direction-dependent schedule decomposes into two independent design choices: a variance path, which fully determines the intermediate laws and probability flow, and a factorization, which leaves this flow unchanged while controlling irreducible regression variance. On the sampling side, we analyze finite-step Euler accuracy and derive a necessary drift bound for exact N -step sampling, connecting the geodesic and the logarithmic path. On the training side, for any fixed path, we derive closed-form factorizations that either minimize time-averaged regression variance or make it constant along the path.

Cancer progression often requires multiple driver mutations, but the same drivers may be acquired in different orders. How these pathways jointly shape tumor clonal structure remains unclear. We develop a multitype branching-process model in which malignant transformation requires two driver mutations, distinguishing malignant cells by mutation order and the independent transformation event that founded their clone. Under a successive exponential approximation, we establish point-process limits for pathway-specific clone sizes and derive a closed-form expression for the limiting expected Simpson's index of the combined malignant population. When both mutation orders yield malignant cells with the same net growth rate, the index decomposes into effective pathway weights, determined by mutation rates and birth-death dynamics at preceding stages, and within-pathway concentration terms, determined by intermediate-to-malignant growth-rate ratios. A driver's effect on heterogeneity thus depends critically on when it is acquired. A strong driver acquired early expands the intermediate lineage and increases the supply of independent malignant founders, whereas the same driver acquired last strengthens the growth and age advantage of early-founded malignant clones. Under additive fitness effects, these opposing mechanisms can produce a non-monotone relationship between selective advantage and clonal concentration. Threshold-like non-additive fitness effects can generate highly concentrated malignant populations, while order-dependent terminal fitness causes the faster-growing pathway to dominate asymptotically. These results show how mutation order, mutational accessibility, selection, and epistasis jointly determine lineage-level intratumor heterogeneity.

This work presents a video summarization algorithm based on multi-object tracking and person reidentification. We integrate facial embeddings, 3D body-shape features, and visual appearance into a unified tracking framework. These representations enable hierarchical identity assignment and tracking through bidirectional anchoring, which robustly recovers trajectories under severe occlusion or low visual quality. From these stable trajectories, we generate a compact set of summaries for each identity. We select keyframes using a multi-factor weighting scheme that optimizes biometric clarity, social interaction, and motion dynamics, while Adaptive Non-Maximum Suppression ensures temporal diversity. Evaluation on a custom dataset demonstrates tracking stability, achieving an IDF1 of 97.89% and a MOTA of 95.79%. Compared to Top-K selection, our algorithm also increases visual diversity by 146%, temporal coverage by 89%, and information retrievability by 3.5%.

Multi-task optimization (MTO) addresses a set of optimization tasks simultaneously, often suffering from inaccurate inter-task relationship estimation under limited evaluation budgets, leading to negative transfer. This paper introduces In-Context Guidance Multitask Optimization (ICG-MTO), a novel framework that leverages numerical foundational models to improve inter-task coupling estimation in few-shot scenarios. Unlike conventional methods that rely solely on scarce observed data, ICG-MTO employs a frozen foundational model to infer auxiliary guidance through in-context learning. The framework operates through three stages: constructing an algorithm-specific in-context query from evaluated solutions, using the foundational model to infer a guidance signal characterizing predictive relationships among tasks, and translating this signal into algorithm-specific guidance for maximum-a-posteriori coupling estimation. This approach provides regularization during the early, data-scarce stages of optimization and gradually relinquishes control as task-specific observations accumulate. We instantiate the framework in multitask Bayesian optimization as ICG-MTBO, using directional fitness-class queries to guide inter-task coupling estimation, and further instantiate it in MFEA-II using decision-space-overlap queries to guide random mating probability estimation. Experiments across synthetic benchmarks and a real-world robot arm control problem, together with evaluations under different acquisition functions and evolutionary multitasking, demonstrate the effectiveness and generality of ICG-MTO for few-shot multitask optimization.

Indoor positioning benchmarks rarely co-locate multiple technologies with high-accuracy ground truth, limiting cross-method and cross-technology comparison. We evaluate four range-based positioning methods on fused Ultra-wideband, Bluetooth Low Energy, and WiFi ranges recorded at 48 reference points in an industrial facility with available ground truth. The evaluated methods range from least-squares positioning, with and without robust weighting, through a Bayesian grid filter to a ResNet regressor that we report under two protocols, interpolation and spatial generalisation. Per-technology ranging quality spans roughly two orders of magnitude, which caps any geometry-based solver that weighs anchors equivalently. On fused input the grid filter is competitive with deep learning on median error, while the ResNet's advantage concentrates on the upper tail. Holding reference points out of training increases the learned regressor's median error several-fold, exposing a spatial-generalisation penalty invisible under random splits. The dataset and evaluation code are publicly available.

We introduce a nullspace-guided adaptive (NGA) bootstrap method that improves the energy lower bounds of quantum many-body ground states by refining the bootstrap basis in a dynamic and incremental way. At each iteration, the optimized moment matrix reveals a nullspace of saturated positivity directions, which is intuitively interpreted as annihilators of the approximate ground-state subspace. The NGA bootstrap then prunes operators with small nullspace leverage and grows the basis along descendants of these null directions. By applying the NGA bootstrap to the transverse-field Ising chain, we obtain nearly exact energy lower bounds because the algorithm automatically discovers the eigenoperator structure in terms of Jordan-Wigner fermions from a minimal local bootstrap basis. For the Hubbard chain, it improves upon state-of-the-art energy lower bounds by up to two orders of magnitude, reaching errors ranging from $10^{-3}$ down to $10^{-5}$ in the strongly correlated regimes. We further show that the NGA framework can be used to improve the certified two-sided bounds on general observables. In addition, the bootstrap error decreases approximately as a power law with increasing computational resources. These results suggest that our method provides a practical and scalable route toward accurate bootstrap of general quantum many-body systems.

Automatic fact-checking poses a significant challenge in Arabic natural language processing due to the scarcity of datasets and resources. In this manuscript, we introduce Arafa, a new large-scale dataset for fact-checking in Modern Standard Arabic, constructed through an automated framework leveraging large language models (LLMs). The dataset was constructed through a three-step pipeline: (1) claim generation from Arabic Wikipedia pages with supporting textual evidence, (2) claim mutation to generate challenging counterfactual claims with refuting evidence, and (3) an automatic validation step to validate that the generated claims are either supported or refuted by their accompanying evidence, or if the evidence does not provide enough information to judge the validity of the claims. The resulting dataset comprises 181,976 claim-evidence pairs labeled as supported, refuted, or not enough information. Human evaluation carried out on a test sample from the dataset demonstrated strong inter-annotator agreement (kappa = 0.89) using Cohen's Kappa for supported claims and (kappa = 0.94) for refuted claims. Automatic validation based on a human-evaluated sample achieved 86% accuracy for supported claims and 88% for refuted ones. To showcase Arafa's value as a resource for automatic Arabic fact-checking, four open-source transformer-based models were fine-tuned using Arafa, with the top-performing model achieving a Macro F1-score of 77% on the test data. In addition to Arafa being the first large-scale dataset for Arabic fact-checking, our framework presents a scalable approach for developing similar resources for other low-resource languages.

Part segmentation is a fundamental problem in computer graphics and 3D vision. Recent works have expanded 3D part segmentation beyond fixed taxonomies, but existing approaches typically only address a specific setting, such as text-guided part segmentation or point-based interaction. In this work, we argue that these settings can be unified as an intent-conditioned generative problem, where different prompts specify the desired part decomposition. To this end, we introduce PartLLM, a unified multimodal model that formulates 3D part segmentation as autoregressive semantic decomposition. Conditioned on an input shape and a user prompt, PartLLM autoregressively generates semantic part hypotheses as queries for mask prediction and feeds them to a decomposition-aware decoder that jointly predicts coherent part masks. This unified design supports text-guided part segmentation, interactive segmentation, and full-shape semantic decomposition with controllable granularity within a single model. Extensive experiments across these task settings show that PartLLM consistently outperforms task-specific baselines, demonstrating the effectiveness of unifying 3D part segmentation under an intent-conditioned generative formulation.

Recent VLM-based autonomous driving planners adopt GRPO-style reinforcement learning to optimize driving performance. However, existing GRPO recipes either optimize driving efficiency, risking progress-seeking but unsafe behavior, or enforce early safety constraints, leading to overly conservative behavior; both require lengthy training. To solve these problems, we first reveal two distinct RL regimes: a progress regime (Run-GRPO) that aggressively explores high progress, and a safety regime (Walk-GRPO) that restores safety under stable progress. Based on this finding, we propose $\textit{Run-then-Walk}$, a simple yet effective two-stage reward scheduling strategy for GRPO, achieving both better performance and faster convergence. Unlike one-stage RL, which may focus on progress, safety, or a mixture of both within a single training phase, this schedule explicitly separates progress discovery from safety repair. In the $\textit{Run}$ phase, we focus on progress, allowing the policy to escape the conservative bias and discover high-progress modes. In the subsequent $\textit{Walk}$ phase, we introduce endpoint and safety strategy to repair unsafe behaviors from the Run phase. This reversed schedule overcomes the conservatism of Walk-first methods and the unsafe progress-seeking of joint optimization. We validate it with various VLM-based planners on multiple benchmarks: NAVSIMv1, NAVSIMv2, Navhard, and nuScenes. Extensive experiments demonstrate improved driving performance while requiring 40–50% fewer RL training epochs than the baselines. Code is available at https://github.com/haha-yuki-haha/AutoDrive-P3_with_Run-then-walk.

Existing waveform-domain audio watermarks are robust to many conventional distortions but can degrade substantially under neural codec resynthesis. We investigate whether continuous neural codec latents provide a more suitable embedding space using a restricted formulation built around frozen pretrained EnCodec. To test this, a feedforward embedder maps a multi-bit payload to an additive latent perturbation decoded through the unchanged codec decoder. Compared with AudioSeal and WavMark, our latent watermark formulation degrades more gradually under repeated and low-bitrate EnCodec resynthesis, transfers to unseen DAC, and retains high detection under most waveform distortions. Substantial EnCodec robustness emerges even without codec-resynthesis supervision, indicating that this behavior is inherent to our latent formulation and is further strengthened by codec-aware training. Learned perturbations are also preserved more strongly through codec cycling than equal-norm random controls, with preservation depending more on channel-specific allocation than temporal structure. End-to-end perceptual quality remains close to that of the frozen EnCodec reconstruction, indicating that much of the observed degradation originates from the codec carrier itself. Overall, these results show that continuous neural codec latents provide a promising embedding space for watermarks that remain robust to neural codec resynthesis.

Cavity-induced electron mass renormalization is an important ingredient of many proposals to control material properties through vacuum field modification. What is its ultimate limit? Here this question is answered by (i) noting that this mass renormalization is essentially due to the field-mediated interaction between the electron and the polarizable degrees of freedom in the cavity mirrors, and (ii) directly evaluating the corresponding Feynman path integral for a 2D electron by numerically exact quantum Monte Carlo. The sought upper limit is provided by the Landau-Pekar polaron in the planar geometry. The presented calculation bridges the gap between cavity control and dielectric engineering of material properties.

The tensor product multilevel method (TPML) approximates high-dimensional functions from scattered data on general bounded domains by combining Smolyak's sparse grid construction with the kernel-based multilevel interpolation, coupling the kernel scale to the fill distance at each level so that condition numbers stay uniformly bounded. Its original formulation depended only implicitly on the target function, which precluded numerical evaluation. Using a recent nodal representation of the low-dimensional multilevel operators, we derive an explicit, computable representation of the high-dimensional interpolant. On this basis we prove error estimates in mixed-regularity Sobolev and continuity norms, extending the existing $ L_2 $-theory, and we characterise the full range of anisotropy weights that attain the optimal rate against the number of degrees of freedom. Three explicit weight strategies follow, equilibrating the accuracy, the degrees of freedom, or the cost-benefit ratio. Numerical experiments in up to ten dimensions illustrate the theory.

Reported gains in vessel-trajectory prediction from Automatic Identification System (AIS) data are credited to new architectures, but the evaluation protocol is rarely measured as a source of error reduction. We build a leakage-aware protocol with vessel-, time- and region-disjoint splits and apply it to two corpora with different traffic: 31 days of Danish national AIS traffic and 30 days of US Gulf coast traffic off Houston and Galveston. On both, we audit TrAISformer, GATransformer, and controlled AISFormer-inspired reconstructions. Three protocol effects appear in both corpora. First, TrAISformer's best-of-16 oracle decoder lowers error by a factor of 2.1-3.2 relative to greedy decoding. Second, a split that shares vessels lowers its greedy error by 23-25% at one hour, against 2% or less for a compact 0.43 M-parameter encoder. Third, a region-disjoint split raises TrAISformer's one-hour error from 2.2 to 24.6 km on the US corpus, because 99.9% of the test contexts fall in longitude bins never seen in training; the encoder built on local offsets is unaffected by this. Architectural mechanisms matter less: GATransformer's graph attention gives no measurable benefit on either corpus, while its waterway feature is worth 12-22%. The effect of a time-disjoint split is not stable across corpora (13% versus 2%). We release the splits and code.

We analyse Schrödinger operators with accretive potentials in weighted spaces. We find conditions on potentials and weights for which the Dirichlet realisation, introduced by generalised form methods, has non-empty resolvent set. We establish a domain and graph norm separation property, as well as sufficient conditions for the compactness and Schatten class of the resolvent. Moreover, we investigate the relation between discrete spectra and eigenfunctions of operators in standard and weighted spaces. As applications we extend results on the completeness of eigensystems of operators with accretive potentials from standard to weighted spaces and analyse operator matrices exhibiting a Schur dominance property, in particular, related to a wave equation with strong accretive damping.

Personalised search must satisfy query intent while incorporating user context and historical interactions. LLM-based cross-encoders provide a single reranking interface, but injecting predictive behavioural statistics into their prompts can encourage shortcut learning: reliance on historical signals at the expense of semantic and user-context patterns that generalise to sparse or unseen searches. We study this problem in the personalised search system of a large-scale audio streaming platform using Query Slice Stats (QSS), an interaction-derived behavioural feature summarising historical success for query-candidate pairs. Naive QSS injection improves ranking when the feature is available but reduces robustness when it is removed. We address this with deterministic dual-sample feature-dropout training, which presents each example once with QSS included and once with QSS removed. Offline, QSS injection improves ranking quality by 13.3% when available. Dual-sample training preserves these gains while improving performance under QSS-removed evaluation by 4.0% relative to naive QSS training. In a live online test, both QSS-aware variants improve search success by roughly 2%. The aggregate test does not distinguish dual-sample from features-only training; the cold-start comparison is directionally consistent with the offline results. Paired feature-present and feature-removed training can therefore reduce the tension between exploiting strong behavioural statistics and remaining robust when they are unavailable.

We review Bayesian models for estimating the uncertainty induced by missing higher orders in perturbative predictions. These models return probability distributions, which have peculiar shapes typically very different from the Gaussian distribution often assumed in physics analysis. In this work, we are able to obtain smooth probability distributions with quasi-Gaussians shapes by suitably modifying the likelihood and the priors of the models.

We solve the finite basis problem for all subvarieties of $\mathsf{V}(S^0_7)$. By considering their intersections with $\mathsf{V}(S_7)$, we reduce the problem to six intervals. We show that $\mathsf{V}(S_7^0)$ has exactly fifteen finitely based subvarieties and that every other subvariety is nonfinitely based. Moreover, $\mathsf{V}(S_c(abc))$ and $\mathsf{V}(SR_6)$ are the only limit subvarieties of $\mathsf{V}(S_7^0)$.

Partial audio deepfakes manipulate only selected speech regions, making them difficult to be localized. Existing methods exploit boundary cues for partial deepfake localization, but primarily focus on identifying boundary positions rather than modeling the feature changes that characterize authenticity transitions. Meanwhile, the internal characteristics of continuous bona fide and spoofed segments remain underexplored. In this paper, we propose Boundary and Intra-Segment Learning (BISL), which introduces boundary learning to model feature differences between adjacent frames and distinguish authenticity transitions from general acoustic variations. In addition, intra-segment learning captures the overall characteristics of continuous bona fide and spoofed segments while enhancing feature consistency within each segment. By jointly learning frame, boundary, and segment information, BISL enables more effective fine-grained partial audio deepfake localization. Experiments on multiple localization benchmarks show that BISL achieves an EER of 2.52% and an F1-score of 97.40% on PartialSpoof, outperforming the compared methods, while maintaining competitive performance on HAD and improved cross-dataset performance on LPS. The code will be made publicly available upon acceptance.

Persuasive video generation (PVG) is a valuable yet under-explored research topic. Despite the significant advances in multimodal content generation, AI-empowered automated creation of human-made-like videos with substantial persuasiveness remains a formidable challenge. In this paper, we propose CogenPVG, a novel Cognitive-Enhanced reflective multi-agent framework tailored for Persuasive Video Generation task. Given the topic and stance from the user, we decouple the sophisticated generation process into four sequential stages: argument reasoning, storyboard planning, asset creation, and post-editing, imitating the workflow of human video producers. To ensure high persuasiveness, each stage is equipped with a pair of generator and critic agents, following a reflective refinement scheme grounded in a solid psychological theory of persuasion, the Elaboration Likelihood Model (ELM). In the argument reasoning stage, we generate highly logical and credible reasoning thoughts under the guidance of critical thinking theory, enabling cognitive enhancement via the central route of the ELM. For the other three stages, we generate and optimize multimodal assets, assembling them into a persuasive video guided by theories of heuristics, as the peripheral route of the ELM. To the best of our knowledge, CogenPVG is the first work focused on general persuasive topics, without being confined to commercial purposes. Extensive experiments and comprehensive analysis demonstrate that our framework achieves the best persuasion performance, thereby proving the effectiveness of our proposed multi-agent framework for the PVG task.

Discrete action tokenization is central to autoregressive vision-language-action (VLA) models, yet action representations are often evaluated primarily through reconstruction fidelity. We ask which representation properties actually matter for closed-loop control by comparing fixed analytical, data-driven linear, and nonlinear neural representations under a unified tokenization interface. Across rate-distortion analysis, sequence-modeling diagnostics, and 3,500 LIBERO rollouts, representation rankings change with the evaluation criterion. PCA achieves lower nominal reconstruction error than Temporal-DCT, but produces less predictable token sequences and 3.0 percentage points lower mean seen-task success across three policy-training seeds, with the policy ordering reversing in one seed. In a matched seed-42 ablation, an autoencoder further reduces reconstruction error yet does not yield the strongest policy and exhibits greater sensitivity to discrete token perturbations. These findings show that reconstruction fidelity alone cannot reliably select action representations for autoregressive control, motivating joint evaluation of geometric fidelity, sequence predictability, decoder stability, and closed-loop performance.