Papers

Exploratory reinforcement learning (RL) on an operating bus fleet is impractical,while policies trained only from historical data cannot acquire new experience. Hybrid Offline-and-Online (H2O) RL combines fixed target replay with simulator interaction, but the inexpensive online simulator can differ from the target in transition and event-duration dynamics. We study this cross-fidelity problem for multi-line bus holding and address a failure mode in which lower generalized passenger time coexists with incomplete passenger journeys.

The correlation dimension ($D_2$) provides a scale-dependent characterization of the transition towards homogeneity and a complementary perspective on the clustering of large-scale structure. In this work, we study its full dependence on scales and redshift connecting $D_2$ directly to the matter power spectrum of a given cosmological model, and derive an approximate analytical relation based on an effective cut-off representation of the filtering kernel. We further perform two complementary local (derivative-based) and global (variance-based) sensitivity analyses within the $w_0w_a$CDM model. In general, nonlinear evolution lowers $D_2$ relative to the linear prediction at small scales, with the effect becoming more pronounced at low redshift. Our cut-off approximation, with $k_{\rm cut}=\alpha/r$ and $\alpha\simeq 2.39$, reproduces the exact $D_2$ to better than 1% for the representative cases considered. The sensitivity analysis identifies $\Omega_m$ and $A_s$ as the dominant parameters for the linear correlation dimension and its nonlinear-to-linear ratio, respectively, while $w_0$ and $w_a$ induce scale- and redshift-dependent changes in the nonlinear contribution. These results provide a framework for characterizing the cosmological scale- and redshift-dependent features of $D_2$ and its nonlinear correction.

Open-weight LLMs are released not only as fixed products but also as substrates for downstream fine-tuning. This openness, however, creates legal and ethical risks because users may misuse fine-tuning to instill illicit knowledge or enable hostile operations. Model providers therefore need apre-release defense against such acquisition, motivating the problem of preemptive unlearning. Unlike retrospective unlearning, which removes capabilities already present in a fixed model, preemptive unlearning seeks to prevent their acquisition under unseen attack data and future fine-tuning procedures. Despite its practical importance, this setting remains largely unexplored, presents distinct challenges, and is therefore the central focus of our work. We first verify that existing retrospective methods provide insufficient pre-release protection. Even when forbidden capabilities are suppressed in current outputs, forbidden-domain data can still induce gradients through internal pathways, enabling later acquisition. Motivated by this finding, we propose a gradient-sealing principle that blocks these pathways by pushing relevant pre-activations into the negative region, where ReLU-family activations exhibit zero or near-zero derivatives. Experiments across multiple LLM families demonstrate our stronger resistance to downstream acquisition than retrospective baselines, validating gradient sealing as an effective mechanism for pre-release protection.

Distributed quantum computation connects multiple quantum processing units (QPUs) through quantum communication to jointly perform large-scale quantum computations. Since the number of qubits available at each QPU is limited, it is important to reduce quantum communication while keeping the qubit overhead small. To this end, we study an entanglement-assisted model, in which entanglement consumption serves as a measure of quantum communication cost. For exact deterministic implementations in this model, the operator Schmidt rank of the target bipartite unitary provides a general lower bound on entanglement cost when qubit overhead is unrestricted. However, it has remained unclear how closely this bound can be approached with constant qubit overhead. In this work, we show that this lower bound is attainable for every bipartite Clifford unitary using at most two auxiliary qubits per QPU. In addition, for non-Clifford unitaries specified by exact Clifford+$T$ decompositions with $T$-count $t$, we construct implementations with the same qubit overhead whose entanglement cost is at most $2t$ Bell pairs above this lower bound.

We investigate strong-field orbital dynamics for static spherically symmetric charged bumblebee black holes featuring spontaneous Lorentz-violating effects. Analysis of null and timelike circular geodesics yields the photon-sphere radius, the innermost stable circular orbit (ISCO) radius, and characteristic orbital frequencies for test particles. Within the relativistic precession model for high-frequency quasi-periodic oscillations (HFQPOs), observed twin-peak frequencies are governed by the full set of fundamental parameters $M$, $X=r/M$, $l_1$, $l_2$, and $Q_0/M$. Among them, $l_1$, $l_2$, and $Q_0/M$ enter frequency expressions only in combined forms, giving rise to intrinsic parameter degeneracy. An effective-parameter set $\Theta=(M,X,C,\beta)$ is accordingly introduced to characterize these composite contributions. Bayesian MCMC parameter inference is carried out using HFQPO observational data from three black-hole X-ray binaries: GRO J1655–40, XTE J1550–564, and GRS 1915+105. The 68% credible intervals of composite parameter $C$ all contain the Reissner–Nordström limit $C=1$, revealing no statistically significant net Lorentz-violating correction under our model assumptions. Distinct triples $(l_1,l_2,Q_0/M)$ yield identical QPO predictions, so HFQPO observations alone cannot disentangle Lorentz-violating and charge-related effects; precise constraints on these fundamental quantities require additional, more precise observational data.

Large language models frequently fail to balance staying truthful with being supportive. They often exhibit sycophancy in responses to users, agreeing with false claims, offering unwarranted flattery, and giving advice skewed toward users' expressed views. In reality, sycophancy rarely happens in a single exchange; it may emerge organically as users repeatedly insist or subtly steer the dialogue over time. Current evaluations, however, rely on rigid, single-turn tests or fixed scripts that fail to capture these natural dynamics. Furthermore, these benchmarks often mistake showing basic empathy for yielding, penalizing models for acknowledging a user's feeling. This view may drive future models to over-correct into cold, dismissive rigidity. To address this gap, we introduce FIGS (Factual Integrity and Grounded Support), a dual-axis evaluation framework built around extended, realistic dialogue. We use an adaptive 10-turn conversational simulator that dynamically challenges the target model, reflecting how users repeat requests, push back, or steer a conversation toward a preferred answer. To accurately evaluate these trajectories, we apply a taxonomy that strictly separates Sycophancy (whether the model holds firm to the truth and keeps its praise proportional) from Calibrated Validation (showing empathetic understanding of the user's feelings without overdoing it). We release our complete testing environment, including 500 diverse multi-turn scenarios and an automated judge. Our evaluation of leading models reveals a consistent trade-off: over the course of a sustained interaction, current systems either slowly drift to sycophancy or over-correct into robotic detachment. This demonstrates that balancing honesty with appropriate support throughout a natural conversation remains a critical, unsolved challenge.

Longitudinal beam dynamics simulations are an essential tool for the design and operation of circular and linear accelerators alike, fundamentally supporting everything from RF system design to the interpretation of beam-based measurements. This paper gives a practical, self-contained introduction to longitudinal beam tracking, aimed at newcomers to the field who wish to build, or better understand, their simulation tools. Starting from the analogy between beam tracking and particle-in-cell plasma simulations, the basic equations of motion, their discretisation, and the choice of phase-space coordinates are introduced, followed by the collective effects driven by the machine impedance. The modelling of the RF system and its associated control loops, and the generation and matching of realistic particle distributions are discussed thereafter. Selected topics that require a fully six-dimensional treatment are briefly touched upon, before closing with practical guidance on the design, optimisation, and benchmarking of longitudinal tracking codes.

Quantum state diffusion unravels a Lindblad master equation into stochastic trajectories of pure states. The ensemble of trajectories is described by a probability density on the manifold of pure states, which contains much more information than the density matrix, its first moment. We derive the time reversal of this diffusion. Writing the reversal of a general diffusion on a manifold in Stratonovich form, and passing through the manifold of pure states, we obtain a backward stochastic differential equation for the state vector, valid for any Hilbert space dimension and any Lindbladian. The backward stochastic equation provides a physical, nonlinear unravelling of the inverse Lindbladian that holds only for the ensemble from which it was constructed. Following score-based generative models, we then use the backward equation to generate a new ensemble close to the original. For the depolarizing channel the unravelling is a Brownian motion on complex projective space, so the uniform distribution of pure states is its stationary law in any dimension. Starting the backward equation from that prior, we bound the error of the resulting distribution compared to the original one in different settings: exact and learned scores, continuous and discrete time.

Let $Z$ be a compact Kähler manifold endowed with a Hamiltonian action of a compact connected Lie group $U$, and let $G\subset U^{\mathbb C}$ be a real compatible subgroup. Using the gradient map $\mu_{\mathfrak p}$ associated with the momentum map, we establish a Calabi-Matsushima type decomposition for the Lie algebra of the stabilizer $G_z$ at a critical point of $f\circ\mu_{\mathfrak p}$, where $f$ is the restriction to $\mathfrak p$ of a suitable $Ad_U$-invariant strictly convex function on $i\mathfrak u$. More precisely, we show that the isotropy algebra decomposes into eigenspaces corresponding to nonnegative eigenvalues of an adjoint endomorphism, with zero eigenspace given by its reductive part. This extends the classical decomposition for the complexified action of $U^{\mathbb C}$ to the real reductive setting. As an application, we prove that the identity component of the compact stabilizer at critical points of $f\circ\mu_{\mathfrak p}$ is a maximal compact subgroup of the identity component of the $G$-stabilizer.

Masked diffusion language models (dLLMs) have shown strong potential for faster inference through parallel token generation when combined with confidence-based samplers. However, recent work has shown that such methods can defer unmasking high-entropy fork positions at which multiple plausible continuations exist. This results in reduced generation diversity, as shown by worse pass@k scaling, and limits gains obtainable from RL post-training. To avoid this flexibility trap, prior work advocated for autoregressive (AR) sampling. Here, we show that discarding confidence-based sampling is unnecessary and, once inference cost is taken into account, wasteful. We first propose Fork-dLLM, a simple hybrid sampler that uses AR-style ordering only at uncertain fallback steps while retaining parallel generation otherwise. We then extend the same principle to post-training with ForkGRPO, which uses Fork-dLLM rollouts and applies the GRPO objective only at fallback steps, preserving exact policy-likelihood ratios while substantially reducing rollout and optimization cost. In our experiments, Fork-dLLM matches the strong pass@k scaling of AR sampling while being 2-3x more efficient, and ForkGRPO achieves downstream performance comparable to or better than AR-based GRPO baselines at a substantially lower training cost.

Large-scale events such as political scandals or misinformation campaigns can abruptly shift the collective opinion of a population and reshape the dynamics in ways that standard local-interaction models of sociophysics do not capture. Here, we investigate the dynamics of opinion formation when a group of interacting agents is subjected to sudden, macroscopic exogenous shocks. Our framework is general and applies to arbitrary opinion dynamics models; to illustrate its scope, we consider the voter model and variations thereof. We verify that the mathematical description is excellent in describing the first-passage properties to consensus of the voter model on the complete graph and on complex networks. Remarkably, we identify the conditions under which shocks can be advantageous or detrimental to reach consensus. For the case of the all-to-all noisy voter model under exogenous events, we unveil a rich phase diagram characterized by novel trimodal and asymmetric bimodal states. Ultimately, this work provides a rigorous yet flexible stochastic-process framework for understanding the out-of-equilibrium behavior of opinion dynamics under recurrent exogenous stressors.

Inspired by results in quantum information, we report the development of a formalism for optical sensing which abstracts the details of specific optical systems and allows us to form general bounds on measurement based only on the quantity to be measured. We find a universal bound that controls the performance of any passive optical system regardless of the technology used to construct it and that this bound is attained by a unique measurement scheme for each quantity of interest. Shockingly, our optical formalism is nearly identical to the quantum formalism, with essentially the same notions of the state to be measured and the operation of the measurement, along with an accompanying uncertainty principle.

We construct exact representations of finite element functions of arbitrary polynomial degree on simplicial meshes in any spatial dimension by two-hidden-layer multilayer perceptrons (MLPs). The first hidden layer uses $\mathrm{ReLU}^0+\mathrm{ReLU}$, and the second uses the degree-dependent activation $\mathrm{ReLU}^k$. The neural networks can realize both zero-skeleton and prescribed-skeleton representatives of discontinuous finite element functions, where the former vanish on the mesh skeleton and the latter admit independently prescribed polynomial data on each relatively open lower-dimensional mesh subsimplex; in particular, continuous finite element functions can be represented pointwise on the closed domain. In the continuous case, a half-open decomposition reduces the second-hidden-layer width by assigning each subsimplex to a single incident element. The number of nonzero parameters grows linearly with the number of elements for fixed spatial dimension, polynomial degree, and output dimension, and it is of the same order as the number of global degrees of freedom of the finite element space in the continuous case, provided that the mesh is shape-regular. All parameters in the neural network can be explicitly computed via the information of the mesh and finite element function without training.

We study the width required for a randomly initialized hidden layer of a neural network to achieve rank lifting. Namely, given a dataset $X \in \mathbb{R}^{m \times d}$ of $m$, $d$-dimensional input vectors separated by an angle of at least $\theta$, we consider the random feature matrix $\sigma(XR)$, where $R$ is standard Gaussian. For positively homogeneous nonpolynomial activations, which include sign, Heaviside, ReLU, and ReLU powers among others, we prove that $$n \gtrsim \frac{1}{\theta}\max\left\{m,\log\left(\frac{1}{\delta}\right)\right\}$$ neurons suffice for $\sigma(XR)$ to have full row rank $m$ with probability at least $1-\delta$. This dimension-free bound exponentially improves the previous general-dimensional guarantee for sign features (Drago et al., 2026) and is essentially tight. The proof shows that one random feature column escapes every proper subspace of $\mathbb{R}^m$ with probability $\Omega(\theta)$, using a coupling of nearby Gaussian directions and a local crossing of the induced hyperplane arrangement. We also study stable rank lifting, where the goal is to establish a quantitative analogue of exact rank lifting, i.e., a lower bound on the smallest eigenvalue of the empirical feature Gram matrix in high-probability. Our analysis unifies and generalizes stable rank guarantees for all $q$-homogeneous non-polynomial activations following prior work in Panigrahi et al. (2020) and Song (2026). In particular, we combine a diagonally dominant Taylor tail of the population kernel with truncation and matrix concentration, to show that for positively homogeneous nonpolynomial activations, stable rank lifting is achieved at width $$n \gtrsim C^q \frac{m}{\theta^{2q+1}} \log^{2q+\frac{1}{2}}\left(\frac{m}{\theta}\right) \log\left(\frac{m}{\delta}\right),$$ where $q$ is the degree of the activation and $C > 0$ is some universal constant.

In this paper, we present an approach for combining stochastic nonlinear model predictive control (SNMPC) and reinforcement learning (RL) to enable probabilistically-safe perception-based navigation in unknown environments. Our method first uses RL to train probabilistic actor-critic and sensor prediction models. We then leverage these probabilistic models in a sampling-based SNMPC framework known as Probably Approximately Correct (PAC)-NMPC, which uses hard constraints to enforce finite-time statistical guarantees on the probability of collision and value function improvement. By ensuring that our finite-horizon SNMPC policies decrease the value function in expectation, we can approach the long-horizon performance of the RL approach while satisfying probabilistic safety constraints. Through simulation experiments, we show that our approach can improve the safety of perception-based RL navigation policies and scale to high dimensional systems with large sensor input spaces and complex nonlinear dynamics. We also demonstrate our approach through hardware experiments, showing improved performance for vision-based navigation with an agile fixed-wing aerial vehicle in unknown environments.

Role-playing prompting has become a popular yet simple technique for improving LLM reasoning and output quality. However, whether it consistently boosts performance across diverse domains remains unclear, as systematic validation is lacking. To fill this gap, we run multi-model, cross-domain, and multilingual experiments on MMLU and MMLU-Redux. We find that gains from role-play prompting depend heavily on model capacity, knowledge domain, and prompt language. Drawing on metacognition theory, we propose the persona-related cognitive alignment hypothesis: role-play works only when the LLM correctly grasps the designated persona and its associated knowledge domain. We test this hypothesis through persona information richness ablation, layer-wise entropy divergence analysis, and latent thought-space deflection observation. To reduce persona cognitive bias and stabilize role-play performance, we propose Mixed-Language Concatenate Prediction (MLCP), a simple, training-free, and efficient multilingual prompt concatenation strategy. It aggregates semantically equivalent role prompts to enrich complementary representational cues. Extensive experiments show that MLCP consistently outperforms vanilla role-play prompting across all tested LLMs.

Pitch tracking systems typically couple a per-frame fundamental frequency ($F_0$) estimator with a temporal smoothing stage to obtain continuous trajectories. Conventional Viterbi smoothers enforce first-order continuity but lack long-term temporal awareness and could lock into octave errors across corrupted frames. We propose Relative Interval Networks (RIN), a trajectory smoothing framework that reconciles per-frame pitch estimates with data-driven multi-hop pitch differences. We extract robust relative pitch intervals across arbitrary frame offsets using Variable-Q Transform cross-correlation. We formulate pitch smoothing as an $L_1$-norm optimization problem and prove its equivalence to a minimum cost circulation problem, solved efficiently via linear programming. Evaluations across speech, singing, and instrumental datasets show that RIN substantially improves weak estimators, matches or outperforms Viterbi decoding at a comparable computational cost, and provides superior robustness under certain acoustic degradation.

EPR-pair consumption is commonly used as a communication-cost objective in distributed quantum computing, but minimizing EPR cost does not necessarily minimize distributed execution time. Despite its widespread use, the reliability of EPR count as a runtime surrogate has received limited direct characterization across different workloads and communication conditions. This work addresses this gap by systematically evaluating the relationship between EPR cost and runtime over all balanced two-QPU mappings of QFT, QAOA, and CDKM circuits. The results show that mappings with identical EPR cost can have substantially different execution times, lower-EPR mappings can be slower, and minimum-EPR mappings can be runtime-suboptimal. The reliability of EPR count is strongly workload dependent and also changes with the communication operating regime: increasing EPR-generation latency amplifies runtime differences hidden by equal EPR cost, while increased communication concurrency can reduce the ability of EPR count to preserve runtime ordering. These results show that EPR count and execution time are related but distinct optimization objectives, and identify regimes in which explicit communication-aware runtime modeling is necessary.

Laser-based deorbiting has long been proposed as a promising technique for mitigating the growing population of orbital debris, particularly decimetre-scale objects – large enough to cause catastrophic collisions yet too small to be economically captured by current active debris removal (ADR) missions. However, the practical implementation of laser-based approaches remains constrained by several factors. Ground-based systems suffer from atmospheric absorption, scattering, and turbulence, which degrade beam quality and limit effective energy delivery. Space-based platforms benefit from superior propagation efficiency, enabling more compact and lower-power systems, but require complex, high-cost orbital infrastructure with demanding deployment, maintenance, and scalability challenges. This paper proposes a stratospheric balloon-based concept for laser debris removal, exploring the intermediate regime between ground-based and space-based approaches. Operating above 99% of the atmospheric mass, such a platform may enable ultraviolet (3HG/4HG) laser operation – largely precluded from ground level by ozone absorption – with improved propagation and ablation coupling efficiency, while presenting a different set of engineering constraints than orbital systems. Rather than a fully developed system, this work focuses on the conceptual design and underlying rationale. We describe the main architectural elements, discuss the expected operational envelope, and position the concept against existing ground- and space-based solutions, identifying potential advantages as well as key challenges and limitations requiring further investigation. Developed within STRATOLASER, an EIC Pathfinder Challenge project targeting TRL 4 through stratospheric balloon experiments.

In this work, we present a simple method to engineer a small neural network to control power converters. Learning-based methods have emerged as a promising approach for power converter control over wide operating point variations. However, many of these approaches do not exploit the analytical structure of converter dynamics and, therefore, search over unnecessarily broad policy classes. We address this gap with a systematic framework that combines model-based trajectory generation with policy learning. First, we use differential flatness of averaged converter models to generate feasible state and constrained duty cycle trajectories where duty acts as a control effort signal. Then for policy learning, we use these model-generated trajectories to construct labeled state-to-duty data and train a multilayer perceptron to approximate the control law with a static map. We apply this framework across buck, boost, and buck-boost converters through topology-dependent flat output selection and a common learning pipeline. Simulations show close agreement between learned and model-generated duty trajectories and stable regulation under large synchronous input and load disturbances.

Foundation models are increasingly adopted across a wide range of applications, often serving as core blocks within AI systems. Yet different foundation models may encode the same input from multiple different views, leading to substantial representation disagreement, which we term Rashomon Representation. Such disagreement often signals inputs that a given model encodes in a way inconsistent with other models, offering a valuable yet underexplored signal for input reliability estimation. While prior work has largely focused on measuring disagreement across multiple models with a representation set, we instead focus on predicting disagreement from a single representation. We hypothesize that this disagreement follows some consistent, input-dependent patterns rather than occurring at random. To test this, we quantify disagreement by comparing each sample's nearest neighbors across different models' representation spaces, then train a lightweight predictor that estimates disagreement from a single model's representation. At inference time, given a new input, the predictor uses that input's representation to tell whether it aligns with or diverges from those of other models. Extensive experiments across diverse foundation models and datasets show that representational disagreement is indeed input-dependent, predictable, and generalizable, enabling efficient reliability estimation of foundation models.

Audio language models (ALMs) are increasingly used for audio deepfake detection (ADD), yet existing benchmarks assess their verdicts or rationale plausibility without verifying the underlying acoustic evidence. To address this issue, we first introduce spoofing evidence-grounded audio reasoning (SEAR), a four-task AQA benchmark to evaluate ALM-based ADD through acoustic evidence identification and quantification, deepfake detection, and forensic rationale generation. We further propose a bona-fide-based acoustic evidence agent (BAEA), which equips a frozen ALM with controlled acoustic tools under fixed or adaptive evidence-acquisition policies. Experiments with six ALMs reveal a clear gap between plausible rationales and verifiable acoustic evidence reasoning, while BAEA-Fixed improves final verdicts and forensic rationales on both evaluation partitions. Controlled interventions further show that misleading evidence degrades both detection and grounding performance.

We investigate the problem of role-capability leakage (RCL), in which a role-prompted reasoning model generates convincing in-role text while continuing to exhibit capabilities on benchmarks that exceed those implied by the assigned role. For example, when a model is prompted to assume the role of a kindergarten student, one might expect its performance on a mathematics benchmark to reflect kindergarten-level ability rather than expert-level proficiency in solving calculus problems. We introduce RoleCapBench, a curriculum-grounded benchmark for evaluating RCL across six educational roles and four assessment levels spanning elementary school through A-level, and use it to evaluate three open-weight reasoning models. We find that although the models can generate stylistically convincing in-role responses, they consistently fail to align their underlying capabilities with their assigned roles. Naive role prompting yields strong role-voice scores of 1.218–1.389 while retaining above-role accuracy of 0.811–0.898. RCL persists across a range of prompting conditions, including prompts that explicitly instruct the model to match the role's capability level. To mitigate this problem, we propose Injection, an inference-time intervention that combines explicit, role-specific capability guidelines with a guiding prefilled response prefix. Injection improves role-capability alignment across models, reducing above-role accuracy by up to 0.562 while preserving in-role accuracy with a marginal drop of less than 0.058 across most models. All artifacts, including scripts and evaluation data, will be released upon acceptance.

Landmark shape spaces arise from configurations of $n$ distinct landmarks with rigid motions factored out and scale normalized. When equipped with sufficiently regular metrics inherited from right-invariant metrics on diffeomorphism groups, the shape spaces inherit geodesic completeness from geodesics of diffeomorphisms. We now answer the corresponding stochastic question if the Riemannian Brownian motion on landmark shape spaces exists for all time and hence rule out initially distinct landmarks colliding when following a Brownian flow. We show that with general classes of metrics, including degenerate metrics with infinitesimal rigid motions in their null space, landmark shape spaces are stochastically complete, thus making the use of Riemannian Brownian motion for modelling shape stochasticity in applied fields well-founded.

Encrypted key exchange (EKE), introduced by Bellovin and Merritt (IEEE S&P 1992), and Masny-Rindal OT, introduced by Masny and Rindal (ACM CCS 2019), are highly-efficient methods for compiling essentially any KEM into advanced cryptographic protocols, namely password-authenticated key exchange (PAKE) and oblivious transfer (OT), by relying only on idealized symmetric-key primitives. They have become leading candidates for practically-implementable PAKE and OT due to (1) their simplicity, (2) their plug-and-play nature, allowing for flexibility in the choice of KEM, and (3) existing proofs of UC-security (in the classical adversarial model). Due to point (2) above, these compilers yield attractive candidates for efficient post-quantum PAKE and OT, especially given the recent post-quantum KEM standardization efforts. This motivates the question of whether the (UC-)security of these compilers translates to the quantum adversarial model. In this work, we show that it does not. In particular, we prove that a general family of (O)EKE protocols, as well as Masny-Rindal OT, are not UC-secure against quantum polynomial-time adversaries, even when instantiated with a post-quantum KEM. To establish UC-insecurity, we devise an adversarial strategy that provably thwarts any attempt by the simulator to extract its input (the password in the case of PAKE, and the receiver's choice bit in the case of OT). To complement these negative results, we establish that both compilers yield certain notions of game-based security. Along the way, we establish a novel “advantage-tight” one-way to hiding lemma that may be of independent interest.

Density estimation is a fundamental problem in statistics and machine learning. In this work, we introduce BayesNDE, a neural density estimator based on Bayesian generative modeling. BayesNDE learns a Bayesian generative model and evaluates its density without requiring invertible networks or Jacobian-determinant computation. For each observation, it infers a sample-specific latent posterior to construct an adaptive proposal that focuses computation on regions contributing most to its density. Bridge sampling then combines samples from this proposal with separate posterior samples to estimate the density. Experiments on nonlinear and multimodal synthetic datasets show improved estimation of density values and better recovery of the density structure compared to the state-of-the-art neural density estimators. Applications to real-world datasets further demonstrate improved anomaly detection. Together, these results highlight BayesNDE as a flexible and effective neural density estimator, demonstrating how posterior inference can turn generative models into tools for density estimation. The code and tutorials are available at https://github.com/liuq-lab/BayesNDE.

GX 5-1 is one of the brightest persistent neutron-star low-mass X-ray binaries in the Galaxy and one of the classical Z sources accreting at luminosities close to or above the Eddington limit. We present the first high-resolution X-ray spectroscopic study of GX 5-1 with the Resolve micro-calorimeter onboard the X-Ray Imaging and Spectroscopy Mission (XRISM), based on a 50 ks AO1 observation. With an energy resolution of approximately 5 eV at 6 keV, Resolve enables a sensitive non-dispersive search for weak and narrow spectral features in this luminous source. During the observation, GX 5-1 was located on the horizontal branch of the Z track. Although rapid spectral variability is detected on timescales of tens of ks, the variability is dominated by changes in the continuum spectral shape. No statistically significant emission or absorption features are detected in the time-averaged spectrum. Using matched-filter searches and time-resolved spectroscopy, we place stringent upper limits on the equivalent widths of low-ionization Fe K-alpha emission and highly ionized Fe XXV and Fe XXVI K-alpha transitions of $EW \lesssim$ a few eV. These results reinforce the picture that, in GX 5-1, classical Fe K diagnostics such as reflection- or wind-related features are either intrinsically weak, strongly ionized, geometrically diluted, or suppressed by rapid continuum variability. Placed in the context of the long observational history of GX 5-1 as a continuum-dominated Z source, the present XRISM result shows that this picture continues to hold even at calorimeter resolution. GX 5-1 therefore provides an important benchmark for high-resolution studies of luminous neutron-star accretion flows and demonstrates the capability of XRISM to place meaningful constraints on line formation even when no significant features are detected.

Reconstructing dynamic driving scenes from recorded sensor data supports closed-loop evaluation of autonomous driving systems by synthesizing observations beyond the original trajectory. Unlike cameras and LiDAR, radar measures radial velocity directly through Doppler. Yet existing radar novel-view synthesis fails to exploit this capability: methods addressing dynamic scenes reconstruct only range-azimuth tensors, while methods that render Doppler assume static scenes. Moreover, because radar processing spreads each reflection across multiple bins, existing representations absorb this spread into scene geometry, causing it to render incorrectly when the viewpoint moves. We present DyRAD, which models dynamic driving scenes using static background reflectors and motion-tracked dynamic point reflectors to render complete range-azimuth-Doppler (RAD) tensors. Reflector velocities are derived from object tracks and projected onto the line of sight, making Doppler both a rendered output and supervision for those tracks. Crucially, we render reflectors through a fixed analytic point-spread function (PSF) derived from the radar's signal-processing chain, preventing sensor-induced spread from being baked into the scene representation. Beyond improving scene reconstruction, this separation also enables zero-shot sensor-configuration transfer, allowing the same reconstructed scene to be rendered under different radar specifications without refitting. We evaluate DyRAD on RADIal, Boreas, and a synthetic benchmark across both on-path poses and displaced viewpoints untested by prior work. On RADIal, DyRAD recovers radar detections in 90.7% of reference-detected objects, compared with 26.9% for the strongest baseline.

To address parameter sensitivity commonly in interior permanent magnet synchronous motor (IPMSM) sensorless control, this paper proposes a parameter-robust control framework by extending the adaptive flux observer from surface-mounted PMSM to salient-pole machines. First, parameter mismatches are interpreted as an equivalent flux vector with $d$-axis and $q$-axis components. Permanent magnet flux, $d$-axis inductance, and resistance mismatches are mapped to the $d$-axis flux component, whereas $q$-axis inductance mismatch is mapped to the $q$-axis component. Then, a scalar adaptive flux update is designed to compensate the $d$-axis flux variation by tracking the equivalent flux magnitude. Since the $q$-axis component rotates the observed flux direction and cannot be eliminated by scalar adaptation, a high-frequency $q$-axis voltage injection method is introduced to identify the $q$-axis inductance. Meanwhile, stability analysis proves bounded equivalent flux magnitude tracking and identifies the residual tangential flux responsible for position error. Finally, experimental results show that the proposed method does not require precise resistance, $d$-axis inductance, or flux, while the $q$-axis inductance is provided by online identification.

Class-Incremental Learning (CIL) aims to continuously learn new classes without forgetting previously acquired knowledge. Parameter-efficient fine-tuning with pre-trained models reduces parameter overhead but can suffer from cumulative interference and suboptimal alignment between inference samples and specialized modules. We propose Dynamic LoRA-Experts and Prototype-Ensemble Matching (DLEPEM), a two-stage rehearsal-free framework. DLEPEM allocates a task-specific LoRA-Expert for each incremental task to reduce cross-task interference, then combines frozen pre-trained-model prototypes with task-adaptive LoRA-Expert prototypes for reliable task-level discrimination. Experiments on standard CIL and Few-Shot CIL benchmarks demonstrate strong performance under the evaluated protocols.