Papers

Long-horizon video world models require persistent memory to preserve scene consistency over extended rollouts. Softmax attention retains the full generation history through a growing KV cache, whereas recurrent linear attention compresses history into fixed-size states with substantially lower memory cost. However, we identify severe long-range forgetting in Gated DeltaNet (GDN), where information from distant but relevant scenes is progressively attenuated by subsequent state updates. To address this limitation, we propose HLA-WM, a training-free hybrid linear-attention framework that combines coarse-grained geometry-guided retrieval with fine-grained recurrent linear-state computation. HLA-WM exploits the affine structure of GDN to cache compact chunk-wise transition summaries, retrieve scene-relevant historical chunks using camera geometry, and recompose them into query-specific recurrent states. On the $60$-second SANA-WM-Bench, HLA-WM improves all six aggregate revisit-consistency and camera-control metrics of the base autoregressive generator without additional training, including a $0.74$ dB PSNR gain and a $28.5\%$ reduction in rotation error. The improvements persist after downstream refinement and generalize to MBench-A, where HLA-WM consistently improves all three revisit-consistency metrics across all four subsets and all evaluated inference modes over $547$ samples. At a $60$-second context, HLA-WM reduces historical-state memory by $12\times$ relative to full KV caching while incurring at most a $1.6\%$ reduction in inference throughput. These results demonstrate that selectively addressable recurrent memory can improve long-range scene recall while preserving the efficiency advantages of GDN. Project page: https://caesarhhh.github.io/hla-wm/

Statistical inference in sequential mediation models requires evaluating the cumulative distribution function (CDF) of the product of three normal coefficient estimators, an operation that entails integrating over a non-convex region with boundary $xyz=v$. While the Delta method provides an instantaneous first-order approximation, its asymptotic variance collapses whenever two or more path coefficients approach zero, producing severe undercoverage near parameter boundaries. Nonparametric bootstrapping avoids gradient collapse but incurs an $O(B \times N)$ computational cost that becomes burdensome in large-scale simulation studies or iterative power analyses. We propose a model-based dimension-reduction algorithm that integrates out the third variable analytically under the trivariate Gaussian distribution for arbitrary mean vectors and positive-definite covariance matrices. Partitioning the $xy$-plane into quadrants isolates the sign change at the coordinate axes, reducing the problem to an adaptive two-dimensional quadrature whose fixed-grid cost is $O(n^2)$ in place of $O(n^3)$. In simulation benchmarks across six parameter regimes, the algorithm achieves a mean absolute error of $3.1 \times 10^{-5}$ relative to a $10^{8}$-sample Monte Carlo reference, evaluating the distribution in under one second and yielding plug-in quantile confidence intervals whose empirical coverage was near or above nominal.

This paper revisits frame-wise saliency for audio moment retrieval (AMR). We show that the frame-wise saliency sequence, conventionally used only as an auxiliary output in DETR-based AMR models, can itself serve as an effective source of moment predictions. We convert the saliency sequence into ranked moments using a simple SED-inspired segmentation rule with no learned parameters, enabling moment retrieval directly from frame-wise temporal information. On the CASTELLA dataset, saliency-based prediction consistently outperforms decoder-based prediction from the same model across all 18 runs of QD-DETR and CG-DETR. For QD-DETR, simply replacing the inference output improves R1@0.7 from 21.0 to 36.1. The advantage remains 7-15 points when the two outputs are evaluated at their independently selected best epochs. The same tendency extends to TaskWeave and UVCOM, whereas TR-DETR shows the opposite behavior, suggesting that how saliency construction may matter. The performance gap is especially pronounced for short moments: for queries whose annotated moments average at most 2 s, R1@0.7 improves from 6.3 to 27.8 with QD-DETR. Decoder supervision nevertheless benefits saliency-based prediction, indicating that its role during training differs from the utility of its inference output.

We construct a continuous function $u$ on the closed unit ball in $\mathbb{R}^4$ and a smooth autonomous Hamiltonian $H:\mathbb{R}^4\to[1,2]$ for which $u$ solves $H(Du)=c$ in the unit ball in the viscosity sense for every $c\in[1,2]$. This reveals a striking and distinctive pathology of viscosity solutions: a fixed function and a fixed Hamiltonian need not determine the right-hand side of the equation.

We develop a transfer learning framework for high-dimensional sparse canonical correlation analysis (CCA) with multiple heterogeneous source datasets. Our goal is to improve estimation of the target canonical coefficient matrices by borrowing information from related sources. We characterize source-target similarity through discrepancies in the joint covariance structure and establish minimax optimal rates under both joint and separate prediction losses, explicitly quantifying the gain from enlarged effective sample size and the cost of heterogeneity. To attain these rates, we first construct an oracle transfer estimator based on debiasing and truncated normalization. We then develop a computationally efficient procedure that adaptively achieves the minimax rate without prior knowledge of the sparsity levels, using aggregation to control transfer bias and a truncated normalization scheme to preserve transfer gains under the scale-sensitive prediction loss. A data-driven source detection procedure is further proposed to identify potentially informative auxiliary datasets. Our analysis also yields a generalized sin-theta perturbation theorem that removes a commonly imposed comparability condition on leading canonical correlations in the sparse CCA problem.

Online control of execution speed is essential for deploying robot policies in real-world scenarios, as robots may need to speed up under time constraints or slow down to facilitate human interaction and improve safety. However, imitation-learned policies inherit the execution speed of their demonstrations, and test-time speed modification can introduce unrecoverable out-of-distribution observations, reducing task success. We observe that the directional inconsistency of action chunks reflects task-phase criticality, indicating how aggressively action step lengths can be modified while preserving task success. Based on this observation, we introduce FreeSpeed, a training-free module that post-processes action chunks from pretrained policies. FreeSpeed resamples each predicted chunk at the requested rate, then uses directional inconsistency between adjacent actions as the primary signal for rescaling. This signal adaptively determines how closely the execution speed can approach the requested speed, allowing flexible speed adjustment within the evaluated limits without compromising task success. Across three policy families and 50 simulated tasks, FreeSpeed supports online speed changes, with realized execution rates spanning 0.22x to 2.53x among settings that preserve per-task success. Across four real-world manipulation tasks, FreeSpeed achieves an average success rate of 94.0%, matching the frozen policy's 93.8%, while realizing execution rates from 0.38x to 1.97x.

Autonomous social navigation requires balancing efficiency, physical safety, and social compliance. Reinforcement Learning (RL) methods provide a viable and effective solution but often rely on unrealistic assumptions, such as the knowledge of humans' position and velocity. In this paper, we introduce JESSI (JAX-based E2E Safe Social Interpretable navigation), a lightweight end-to-end RL framework that maps raw LiDAR scans directly to kinematically feasible control commands. JESSI enhances safety via Dirichlet-parameterized continuous action spaces and deterministic bounding, while an integrated attention-based perception module extracts probabilistic human states for interpretable, socially aware decision-making. Through extensive simulations and real-world deployment on a differential-drive robot, we demonstrate that jointly optimizing the RL policy with a supervised perception signal in a multi-task paradigm enhances social behavior. Ultimately, JESSI is able to balance high navigation success rates and superior social behaviors compared to state-of-the-art baselines.

LLM agents rely on long-term memory to retain and reuse information when performing tasks over long horizons. Existing methods provide limited support for handling memories that become outdated as new observations or domain evidence arrive. Such outdated memories may remain semantically relevant, continue to affect dependent records, and retain value as historical evidence. This calls for two capabilities: dependency tracking to identify downstream effects and historical preservation to retain useful past records. We propose Provenance-Aware Cascading Memory Invalidation (PACMI), a framework that represents memories and new evidence in a provenance graph with typed dependency edges. PACMI assigns records to a four-state validity lattice, propagates validity changes to dependent memories, and uses the resulting states for retrieval and stale-premise detection. We also introduce a diagnostic benchmark with 100 cases and 300 queries across five domains. The evaluation separates node, context-, and answer-level performance. PACMI achieves the highest final-answer accuracy on this benchmark, and its paired difference from the strongest baseline is significant under an exact McNemar test. The premise checker achieves perfect precision, recall, and F 1 on the controlled query distribution. Cascading propagation primarily improves memorystate correctness: removing it increases final-answer errors from 3 to 11, but the paired difference does not reach the 0.05 significance threshold. Code and data will be made publicly available.

Earth observation (EO) data provide rich temporal supervision, yet existing remote sensing foundation models mainly exploit sequential observations through imposing predefined pairwise relations or aggregating holistic reconstruction context. We seek to further exploit the sparse and nonuniform temporal sampling inherent in EO sequences as supervisory signals. To this end, we propose T-JEPA, a temporal joint-embedding predictive architecture that learns time-gap-conditioned latent transitions. A shared single-frame encoder processes each observation, while a temporal predictor estimates the complete target latent field from a masked source latent representation and the actual elapsed time. Across multiple temporal intervals, these predictive constraints organize observed states into structured latent trajectories. Asymmetric metadata injection mitigates shortcut learning, and direct supervision across multiple temporal scales proves more effective than recursively rolling out intermediate states. In parallel, masked pixel reconstruction provides complementary supervision for preserving spatial details. Under matched pre-training data and throughput, T-JEPA achieves leading transfer performance on both static and temporal tasks. Analyses further reveal that T-JEPA learns representations with time-gap-dependent transition predictability and coherent latent dynamics, while maintaining strong cross-period consistency, representation diversity, and semantic discriminability.

Mean-field games provide a continuum framework for modeling the dynamics of large, interacting populations of non-cooperative agents. This paper studies symmetry-breaking in a finite-horizon, two-choice min-LQG mean-field game in which identical agents with linear stochastic dynamics choose one of two equally desirable terminal destinations, while trading off control effort against social pressure to conform. The model has an odd symmetry between the two destinations and therefore always admits a symmetric, dynamic Nash equilibrium in which the population splits evenly between them, producing a deadlock collective state at final time. Numerical studies have suggested that, as the penalty for social nonconformity increases, this symmetric equilibrium loses stability as a fixed point of an associated scalar, self-consistency map, and asymmetric consensus Nash equilibria emerge, where most agents select the same destination. By analyzing the linearized forward-backward PDE system through the scalar map representation, we provide a proof of this loss of stability of the symmetric equilibrium. Together with the odd symmetry of the map, this implies the existence of symmetry-broken mean-field game equilibria corresponding to consensus on either destination.

A cosmological constant $\Lambda$ in a universe weakly violating the cosmological principle (CP) can mimic a dynamical dark energy (DDE) in a FRW universe, and therefore challenging the recently reported DDE evidences. We test this scenario in the form of $\Lambda$LTB cosmology, against two data combinations. (1) DESI DR2 BAO, Planck CMB, together with any of the four SNe Ia compilations (UNION3, Pantheon+, DES SN5YR, and DES Dovekie) show no preference over DDE or $\Lambda$LTB. (2) The inclusion of the kinematic Sunyaev Zel'dovich (kSZ) effect rules out a large portion of the CP violation parameter space allowed by (1). However, the existence of a $\sim 300$ Mpc$/h$ region with $\sim 10\%$ matter density excess is still allowed by the current data, and therefore remains a viable alternative to DDE. To facilitate further tests, we present analytical expressions of the distance-redshift relation and the kSZ effect in a universe with weak CP violation.

Generative AI (GenAI) can explore large bodies of literature and generate plausible research ideas, but identifying what is missing, understudied, contradictory, or potentially connected does not by itself reveal where theory should advance. We develop a theory-oriented agentic AI system that helps researchers identify potential theorizing opportunities by incorporating established theorizing approaches into literature exploration and evaluation. The system operates through three stages. Stage 1 expands the theoretical search space and constructs a provisional Candidate Knowledge Graph. Stage 2 independently reconstructs what the literature supports through source grounded evidence extraction and theory-state reconstruction. Stage 3 evaluates the reconstructed knowledge state to determine whether an unresolved configuration warrants theory development or another research action and, when theory development is warranted, which theorizing approach is appropriate. We demonstrate the system through an end-to-end analysis of human oversight of agentic AI systems in organizations. The analysis shows that literature gaps alone are insufficient for identifying theoretical opportunities. For example, "transparency to trust" is routed to mechanism-based theorizing because the relationship is repeatedly documented in prior studies, while the generative mechanism explaining how transparency shapes trust remains insufficiently specified. By combining large-scale literature processing, structured knowledge representation, and theorizing-guided diagnosis, the system serves as a theory-oriented research assistant that supports researchers in identifying theoretically meaningful directions for subsequent research.

We examine the role of the Bakamjian–Thomas (BT) construction in achieving current-component independence for pseudoscalar-to-pseudoscalar semileptonic form factors in light-front (LF) dynamics. An exactly solvable covariant Bethe–Salpeter (BS) model provides an off-shell benchmark. Its transverse-current zero-mode contribution is calculated directly from the shrinking nonvalence region and reproduced by an effective operator in the valence convolution. In the BT-based LF quark model, the one-body current kernel is constructed from on-shell constituent spinors, while interactions are encoded in the mass operator and wave functions. In this realization, the $(+,\perp)$ current-component extraction is zero-mode free and is saturated by the on-shell overlap. Using the free constituent invariant masses consistently in the internal projections yields the same $f_-(q^2)$ from the $(+,\perp)$ and $(+,-)$ current-component extractions after integration. We establish this equality analytically for rotationally invariant radial wave functions with the LF Jacobian and the adopted pseudoscalar spin coupling, provided the reduced overlaps converge. The difference from the minus-current projection using physical meson masses defines an interaction-dependent BT completion, distinct from the BS endpoint contribution. A mass-weighted complex transverse contour permits direct evaluation of the Gaussian $D\to K$ overlaps throughout the physical timelike interval, including zero recoil, without a form-factor parametrization or extrapolation. These results establish current-component independence within the specified BT realization and clarify the dynamical dependence of zero-mode assignments.

Large Language Models (LLMs) have shown strong potential in multivariate time series forecasting and anomaly detection. Existing studies predominantly inject temporal information into LLMs via direct numerical tokenization or heuristic textual descriptions. However, LLMs still face difficulty in perceiving the underlying structural patterns of numerical time series, particularly the seasonal and trend components obscured by discrete numerical tokens. To bridge this gap, we propose FreSia, a frequency-aware framework that establishes an effective alignment between the semantic space of LLMs and the frequency space of time series. Specifically, FGPrompt, a Frequency-Guided Prompt mechanism within FreSia, distills the frequency-domain structures of time series and projects them into prompts tailored to the semantic space of LLMs. Furthermore, we introduce a Global-driven Context Learning (GCL) component, which uses a global CLS-driven probe to generate global context to bridge the time-frequency domain gap and fuse the multi-modal information. Experiments on eight forecasting benchmarks show that FreSia achieves average improvements of 13.48% and 8.06% in MSE and MAE, respectively.

We construct arbitrarily large finite sets $A$ of real algebraic integers such that both $|A+A|$ and $|AA|$ are at most $|A|^{1.95835}$. The construction combines truncated ideal-smooth $S$-unit fibres with a coprime additive factor. A tensor-product rank argument, using a two-dimensional local feature at each selected prime ideal, controls the loss in the additive factor, while a direct estimate for an outer parallel body improves the sumset packing bound. The arithmetic input is an unramified pro-$2$ tower over a known degree-ten field, with simultaneous Frobenius cuts controlling the small prime ideals. We use unconditional Tsfasman–Vlăduţ inequalities for the joint class-number–regulator cost. All finite numerical comparisons entering the exponent are certified by outward rational interval arithmetic. No unproved hypothesis is used.

This paper studies data-driven stabilization of unknown discrete-time linear systems using only quantized state measurements that take values in finite sets. The proposed approach consists of two stages. In the controller design stage, we collect quantized data while maintaining a prescribed quantization error bound and use them to formulate a semidefinite program (SDP). We establish a verifiable condition under which any feasible solution to the SDP yields a stabilizing feedback gain, and show that a stabilizing gain can always be obtained when the quantization error is sufficiently small. In the stabilization stage, a Lyapunov-based quantizer update rule is developed to guarantee exponential convergence under quantized state feedback. As a key feature, the number of quantization cells remains finite in both stages and constant during stabilization. Simulation results illustrate the effectiveness of the proposed approach.

Given incomplete measurements of a single physical field in a coupled system with unknown parameters, can we infer its full physical state and identify the underlying parameters? This problem is challenging because multiple coupled fields must be reconstructed simultaneously from limited observations of only one, while the system parameters are unknown. In this work, we propose a machine learning framework for full-field reconstruction and parameter identification of unknown physical systems from sparse observations of a single physical field. Specifically, the cross-attention encoder propagates sparse sensor observations onto a regular grid to construct a sensor-conditioned latent representation, while a Fourier neural operator (FNO) decoder captures global spatial dependencies to reconstruct all coupled physical fields. The network parameters and unknown physical parameters are jointly optimized by minimizing observation losses, governing equation residuals, and boundary/initial condition constraints. The proposed approach is validated on two- and three-dimensional lid-driven cavity flows, a two-dimensional cylinder wake, and a two-dimensional non-ideal magnetohydrodynamics problem, demonstrating the recovery performance of unobserved fields and physical parameters from incomplete observations.

Let $G$ be a simple graph on $n$ vertices, and let $F(G)$ denote the number of its spanning forests. Bencs and Csikvári [Upper bound for the number of spanning forests of regular graphs, European J. Combin. 110 (2023) 103677] proved that every $r$-regular graph $G$ with $r\geq 2$ satisfies $F(G) \leq r^{n}$. They further conjectured that for $r \geq 3$, \[ F(G)^{1/n} \leq \frac{(r - 1)^{r-1}}{(r^2 - 2r - 1)^{r/2-1}}. \] In this paper, we resolve this conjecture in the affirmative.

We investigate the mass spectra and two-body nonleptonic weak decays of ground-state doubly and triply heavy baryons in a nonrelativistic quark-diquark model. To account for the finite size of the diquark, we generalize the convoluted quark-diquark interaction to unequal-mass diquarks using mass-weighted constituent coordinates and distinct pairwise spin-spin couplings. The resulting Schrödinger equations provide the baryon masses and wave functions used consistently in the decay calculation. The predicted masses differ from existing lattice QCD results by less than $1.5\%$, and our calculated mass of $\Omega_{cc}^{+}$ is consistent with recent LHCb observations. The calculated hyperfine splittings are systematically slightly smaller than the lattice values, indicating a residual uncertainty in the spin-dependent interaction. Within the factorization approximation, we evaluate the $c\to s$ and $b\to c$ transitions with pion or kaon emission. At leading nonrelativistic order, the vector contribution vanishes for the $3/2^+\to1/2^+$ transitions, while both vector and axial contributions enter the $3/2^+\to3/2^+$ channels. The $b\to c$ spatial overlaps are substantially smaller than their $c\to s$ counterparts. Together with the Cabibbo-Kobayashi-Maskawa hierarchy, this reduces the corresponding $b \to c$ width by approximately two orders of magnitude. In the $3/2^+\to1/2^+$ charm transitions, the leading $|\bm{k}|^3$ dependence enhances the suppression of kaon emission. In bottom transitions with a larger energy release, the smaller relative phase-space difference brings the ratios close to the scaling determined by the Cabibbo-Kobayashi-Maskawa elements and meson decay constants. These results provide benchmark estimates for future studies of multiply heavy baryons and quantify the role of recoil dynamics within the finite-size quark-diquark framework.

Thermal X-ray pulse-profile modeling depends on the angular and spectral properties of neutron star atmosphere emission. We construct a compact empirical approximation to the specific intensity using eight coefficients for the angular dependence and five for a normalized spectral kernel. The angular prescription is motivated by the formal solution of the radiative-transfer equation, while the spectral normalization preserves a specified bolometric flux. We calibrate this family to numerical atmosphere tables, focusing on return-current-heated and fully ionized hydrogen atmospheres. When incorporated into a semi-analytic pulse-profile model, the approximation reproduces profiles normalized by their phase-averaged flux to within approximately $4\%$ for antipodal hot spots and $5\%$ for non-antipodal hot spots in the tested heated-atmosphere configurations. We also analyze synthetic eXTP observations using different beaming prescriptions to test the conclusion in our previous work based on a linear description of the angular dependence of the emission. The comparisons suggest that the angular dependence of the intensity cannot be simply characterized by a linear function of the cosine value of the emission angle, because its detailed shape and energy dependence will affect the parameter inference. This parameterization can significantly increase the efficiency for analytic and numerical models, and provide a compact framework for examining how atmosphere assumptions affect neutron star pulse profiles and the inferred stellar properties.

Vision-Language-Action (VLA) models serve as unified policies for robotic manipulation, yet their expensive inference forces robots to pause between policy calls, resulting in stop-and-go execution that interrupts smooth motion and prolongs task completion. Extending the action chunk reduces policy calls and hence these pauses, but predicting farther into the future makes long-chunk execution unreliable. To understand where this unreliability arises, we analyze action errors within long chunks and find that they concentrate around transitions between manipulation subskills, growing sharply with chunk length. This suggests the importance of transition timing, i.e., when to switch subskills within a chunk. Motivated by this observation, we introduce RACE (Reliable Action-Chunk Extension), a framework that predicts the transition timing from an auxiliary one-step denoising pass and conditions action generation on it. By learning and conditioning on transition timing, RACE reduces errors at subskill transitions and enables reliable execution of longer action chunks. Across simulation benchmarks, RACE outperforms fine-tuning at the same chunk length; with 2x longer chunks, it surpasses recent state-of-the-art and efficient VLAs in success rate, and with 4x longer chunks, it remains competitive. On a real robot, RACE uses 4x longer chunks, which reduces the idle time caused by stop-and-go execution by about 5x, while achieving a higher success rate than fine-tuning with the same chunk length. Code and a real-robot demo are available at https://github.com/Seonghoon-Yu/RACE-VLA

We introduce Reduced-rank Mahalanobis Distance (ReMaD), a novel prototypical distance-based refinement to classification and out-of-distribution (OOD) detection using pretrained models without finetuning. We use embeddings of the target dataset to fit closed-form distribution statistics in the model's latent space which can classify in-distribution samples and detect OOD samples, all without training or prior knowledge of the OOD data. Building on prototype classification and OOD detection, we analyze the distribution properties of large pretrained models when processing new datasets; based on this analysis, we formulate a simple modification to Mahalanobis Distance to adapt models' latent space distributions to new domains by removing unused features, without the finetuning or hyperparameter searches required by other adaptation procedures. We demonstrate the efficacy of this method to adapt existing large pretrained image embedding models to new classification domains outside their trained capabilities by testing across four target datasets, with competitive performance in both classification and OOD detection.

In-context retrieval (ICR) is a retrieval-based form of in-context learning (ICL) in which demonstrations are retrieved from a source database based on similarity to the query, rather than sampled independently. In this work, we formulate ICR as a type of domain adaptation problem, where the source distribution $P$ of the database may differ from the target distribution $Q$ of the test query-label pair. We investigate the performance of ICR under a flexible class of distributional shifts that substantially extends prior work \citep{li2024fine,guo2025retrieval}, and establish theoretical guarantees that quantify the benefits and pitfalls of this learning paradigm. Our theory is verified by experiments on synthetic and language tasks.

It has been known that (the square of) Jacobi elliptic function can be expressed by an infinite sum of single solitons of $\sech^2$ shape. In this paper, we show that there exists a similar structure for higher genus cases. It turns out that this is just a consequence of the quasi-periodicity of the Riemann $\theta$-function. More precisely, we show that the second derivative of $\log \theta$ with the real Riemann $\theta$-function of genus $g$ near soliton limit can be well approximated by the sum of real and regular $g$-soliton solutions of Hirota-type in $g$-dimensional real space $\R^g$. This leads to a tessellation of $\R^g$, whose tile is an oblique prism divided into $2^g$ sections by hyper-planes of dominant exponents in the theta function. We apply the results to study quasi-periodic solutions to the KdV and KP equations. We construct the quasi-periodic solutions using the Schottky group, which uniformizes the corresponding Riemann surfaces. We also discuss solitons on quasi-periodic background by pinching some of the homological cycles

Vision-language models (VLMs) can detect that an object has rotated across views, but cannot reliably tell by how much. We introduce OR-Bench, a fine-grained benchmark for object-rotation reasoning with eight tasks covering rotation detection, rotation magnitude estimation, and multi-view rotation reasoning. Across 12 VLMs, the gap is stark: the strongest models approach 100% accuracy on detection, yet even coarse magnitude estimation is near chance. When asked for exact angles, models place 91.8–100% of their predictions on just $0^\circ$, $90^\circ$, and $180^\circ$, a failure we term canonical-angle collapse. This collapse persists even without visual input. Representation probing shows that missing information is only part of the explanation. Although rotation information becomes less recoverable at finer granularity, substantial coarse-grained information remains, and a simple linear probe outperforms the models' generated answers. This suggests that VLMs underuse rotation information they already encode. We therefore propose RotationCue, a lightweight decoder that recovers coarse rotation information from the VLM's own frozen representations and feeds it back to the model as intermediate textual context. Across three VLMs, RotationCue improves every model–task combination on OR-Bench, raising macro-average accuracy by 7.9–12.6 points while preserving general capabilities.

The emergence of spin-polarized electrons from nonmagnetic chiral molecules and crystals is called Chirality-induced spin selectivity (CISS). A chiral phonon of definite wavevector $q$ can strongly enhance it, but near equilibrium the two helicities $+q$ and $-q$ are equally populated: they are time-reversal partners and their contributions cancel. We show that in a biased chiral conductor the current causes the phonon of one helicity to damp less than the other. The damping asymmetry is odd in the current, odd under $q\rightarrow-q$, invariant under a mirror that reverses the structural handedness, and equal, mode for mode, to the current-induced nonconservative Berry force on the phonon coordinate. Closing the phonon kinetics with a rate equation turns the asymmetry into a net lattice helicity: at any finite bias one helicity is preferentially populated, with no threshold, growing linearly with the current as $V\rightarrow0$; above a threshold that helicity becomes a self-sustained coherent travelling wave. We demonstrate this effect in elemental tellurium, treated as an open quantum system (electronic Hamiltonian, phonon bath, and electron-phonon coupling) within a quasi-ab initio framework. The electronic Hamiltonian is supplied by a vertex-corrected quasiparticle self-consistent $GW$ (QSGW) potential with spin-orbit coupling; nuclear displacements are modeled with a machine-learned interatomic potential, which yields chiral phonon modes at finite $\pm q_z$ on the $\Gamma$-A line. The perturbation to the electronic Hamiltonian from a nuclear displacement is approximated by a frozen-phonon deformation potential. No optical pump is needed: a bias alone drives the current and selects the phonon helicity; reversing the current reverses the selection.

The nonlinear Hall effect (NLHE) holds significant promise for next-generation electronic device applications, and it is essential to explore and engineer a strong and highly tunable NLHE. In this work, we report large second-order and third-order NLHEs in dual-gated BiSbTeSe2 devices. Both the second-order and third-order NLHE are highly gate-tunable. The second-order NLHE exhibits a higher generation efficiency than those reported in previous topological insulator systems, and it can be stably observed up to room temperature. Meanwhile, the generation efficiency of the third-order NLHE is comparable to previously reported values. The scaling analysis indicates that the second-order NLHE is consistent with contributions from skew scattering and a term compatible with either the Berry curvature dipole or side jump, whereas the third-order NLHE is consistent with contributions from the Berry connection polarizability tensor and higher-order skew scattering. Our work establishes a promising experimental platform for the development of NLHE-based devices.

We introduce a framework for multi-bit text watermarking with security defined directly through $f$-divergence from the base language model distribution. Unlike prior approaches that focus on average-key distortion-freeness or a particular statistical distance, our formulation supports general $f$-divergences, including total variation and KL divergence, and enforces the guarantee for each realized key and embedded message. We develop a coding-based watermarking scheme that optimally biases next-token distributions subject to a prescribed divergence budget, and characterize the resulting tradeoff between embedding rate, decoding reliability, and statistical security. Experimentally, we compare our method against prior multi-bit watermarking schemes across modern language models and payload regimes. Our approach achieves substantially lower watermark detectability while maintaining competitive message-recovery performance and generation quality. Our results provide a unified view of secure multi-bit watermarking and recover several commonly used security notions as special cases.

Strong image generation models are conditioned on class labels, aligned to frozen pretrained encoders, or built on separately trained autoencoders. While effective, generation then depends on supervision or pretraining: labels must be annotated, and encoders or autoencoders pretrained for the target domain. We study joint generative and self-supervised representation learning in a single model, enabling self-conditioned generation without labels or pretrained models. This is challenging because the objectives are mismatched: contrastive learning consumes clean augmented views and favors coarse, invariant semantics, while flow matching consumes noisy images and must preserve the fine detail and spatial layout that contrastive learning discards. We propose SCION (Self-conditioned Generation on Self-supervised representation), whose core is a single pixel-space encoder conditioned on the flow timestep and an embedding. For representation learning, this conditioning embedding is a learned global vector shared across images, with the encoder's [CLS] token yielding the semantic representation trained by the contrastive loss. For generative training, the conditioning embedding is the image's own [CLS] representation, while patch tokens pass through a decoder to predict the image. To sample without a reference image at inference, we jointly learn a prior over the embedding. Gradient-norm balancing and stop-gradient mechanisms enable joint optimization in one run. SCION is self-supervised and self-contained, with no labels or pretrained models. On ImageNet 256x256, with the JiT-B recipe and no representation guidance, SCION reaches 8.92 FID, surpassing class-unconditional iREPA, which aligns to pretrained DINOv2 (46.44), and RCG, which conditions on it (14.27). With JiT-L, SCION achieves 5.89 FID without guidance and 3.47 with representation guidance, outperforming RCG with the ADM recipe (6.24).

For two decades, the origin of Chirality-Induced Spin Selectivity (CISS), the spin polarization of electrons by nonmagnetic chiral systems without magnetic fields, has remained unsettled. Here, we establish a single exact rule defining the fundamental conditions for the effect: in nonmagnetic, time-reversal-invariant conductors, all measurable CISS signals are time-reversal odd and vanish near equilibrium, proving that structural chirality alone is insufficient. By evaluating leading theoretical mechanisms across four levels of nonequilibrium transport, from coherent classical driving to fully non-Markovian quantum baths, we verify this selection rule to machine precision. We show that coherent chiral vibrations generate substantial collinear polarization (up to 10 percent across 0.5-6 THz) that reverses with handedness, whereas incoherent vibrations yield under 1 percent, and non-Markovian bath memory further suppresses the signal. The chiral geometry first makes the electronic motion chiral by accumulating orbital angular momentum; spin-orbit coupling (SOC) then converts that into spin. CISS therefore acts as both a spin polarizer and a spin filter, the latter an order of magnitude weaker. The polarization grows with molecular length and then saturates, matching trends reported for DNA and peptides. Reversing the drive converts decaying spin into a handedness-locked charge-current pulse (inverse-CISS). We find that once the system is driven, the polarization magnitude scales with an effective spin-orbit coupling: making heavy atoms and curved light-atom backbones indistinguishable at equivalent effective SOC. This enables us to chart how geometry, driving field, and length separate those routes.