Information Retrieval
Papers
Fresh from arXiv, typeset properly. Highlighted words trade on the market.
All of arXiv
Thursday 5 November
Generative recommendation aims to predict item identifiers from user interaction histories, but explicit control over the properties of recommended items remains underexplored. A key challenge is to connect human-readable concepts to internal features whose interventions effectively change the recommended items. To address this, we propose Steerable Generative Recommender (SteerGR), a framework that uses sparse autoencoders for concept-level control in game discovery. SteerGR proceeds through three stages: (i) Concept Definition and Grounding for defining play concepts with an LLM; (ii) Feature Discovery and Selection for discovering candidate features by their concept associations and selecting one by its steering effect; and (iii) Sparse Feature Intervention for amplifying or suppressing selected features to steer recommendations without retraining the recommender. Extensive experiments with two industrial generative recommendation models on a real-world game dataset show that the selected features successfully steer recommendations toward or away from target concepts. We believe this work takes a first step to extend SAE-based steering from LLMs to industrial generative recommendation.
Machine Learning
TopoSHAP: Shapley for Explainable and Efficient Topological Deep Learning
Topological deep learning (TDL) extends learning on graphs, such as molecules or social networks, to higher-order structures, such as rings in a molecule or groups in a social network. Explaining which of these structures drive a model's predictions, and which architectural components matter most for its performance, remains challenging. We introduce TopoSHAP, an efficient and general explanation framework that addresses both questions across topological domains and model architectures. TopoSHAP-cell assigns Shapley values to topological data cells and extracts compact explanations. These explanations match or outperform graph-based counterparts on GraphXAI benchmark data while running up to 112$\times$ faster. They also reveal advantageous topological feature design, leading to measured improvements in explanation recovery and predictive accuracy. On the architecture side, TopoSHAP-neighbor explains how neighborhoods contribute to model performance, and turns these evaluations into a backward-elimination search that identifies smaller, competitive architectures early in training. This is up to $11.3\times$ cheaper than traditional TDL training sweeps.
Read the full abstract
Machine Learning
Not All Latents Are Equal: Importance-Weighted Latent Graph Generation
Graph generation has shown tremendous effectiveness in modeling complex scientific tasks ranging from molecule generation to optimized circuit design. Most recent advancements have developed models that learn to denoise noisy graphs using either discrete diffusion or flow matching. However, discrete graph diffusion scales quadratically with the number of nodes, making it difficult to use such models on larger graphs. On the other hand, latent graph generation instead seeks to denoise noisy graph latents, which scale better to more nodes due to their lower dimension. However, in practice, latent graph generation models lag behind their discrete counterparts in performance. We study this problem, finding that it is caused by the unequal distribution of that different latent dimensions have on graph reconstruction. the impact. We observe that the vast majority of impact is contained in only a smaller number of latent dimensions. Since equal emphasis is placed on reconstructing the entire latent in the loss function, existing models are optimized to reconstruct information irrelevant to the downstream task. To tackle this problem, we propose Latent Importance Reweighting (LIR) which weights each latent dimensions via it's importance to final graph reconstruction. Importance is calculated as the relative influence of a single latent unit on graph reconstruction, which is represented via the Jacobian of the decoder relative to the latent. Our method is model-agnostic and requires no additional model parameters or hyperparameters. Experiments show that LIR can boost performance and lead to faster training convergence when applied to two different autoencoders on various benchmark datasets. Code: https://github.com/biraaj/GraphGenLatentReweight
Read the full abstract
Machine Learning
Training Stability under Adaptive Curricula
Unsupervised Environment Design (UED) approaches train reinforcement learning (RL) agents on diverse curricula that may evolve with their capabilities. While prior work has largely focused on how these curricula are constructed, we study whether their students remain trainable under the resulting distributions. Across multiple UED mechanisms, environment complexities, and extended training horizons, we identify recurring learner-side pathologies, including parameter growth, degraded value–outcome ordering, and loss of representational diversity, that accompany stagnation and performance degradation. These findings connect limitations in UED to optimization failures previously identified in conventional deep RL, while showing how adaptive curricula couple the learner's optimization state to its future training distribution. Controlled interventions show that mitigating selected pathologies improves learning. Stabilizing the learner improves long-horizon performance and zero-shot generalization without changing the curriculum rule or increasing environment interaction. Our results establish learner trainability as a complementary design principle for UED. Effective curricula require students that remain capable of learning from the distributions they produce.
Read the full abstract
Mathematical Physics
Talanov dynamics and chirp-controlled rogue-wave formation for the vector nonlinear Schrödinger equation
We study the focusing vector nonlinear Schrödinger equation, or Manakov system, for a pair of coupled initial conditions with compactly supported, concave parabolic density profiles and phases depending quadratically on the spatial coordinate. We focus on the natural class of self-similar solutions generated by such profiles, whose dispersionless evolution we refer to as Talanov dynamics. We extend the criterion for blow-up in finite time, first established for a single field, to the Manakov case. For arbitrary initial chirps in the two fields, we provide closed-form expressions for their evolution, and an explicit expression for the time of the gradient catastrophe, if it occurs, and we classify the effect of the coupling on the catastrophe. Depending on the initial chirps, the two fields may catastrophise, or relax, as each would if decoupled. However, coupling can also force the catastrophe in fields that would otherwise relax if decoupled. For identical initial supports, and in the presence of dispersion, we numerically show that the same dispersionless criterion determines, in the semiclassical regime, whether the fully dispersive solution relaxes or develops a rogue wave in each field, whose onset time is accurately predicted by the corresponding dispersionless catastrophe time, in direct analogy with the scalar case. As chirp and amplitude are relatively easy to control experimentally, we expect these results to be relevant to experiments in nonlinear optics and Bose-Einstein condensates, extending to coupled systems observations made for the scalar equation.
Read the full abstract
Human-Computer Interaction
Learning from How Users Interpret: Personalizing Recommendation Explanation Order
Personalized explanations typically adapt their content or presentation based on preferences, interaction histories, or other user characteristics. We examine whether regularities inferred from prior textual interpretations can support personalization of explanation organization. In a controlled within participant study with 122 participants, prior interpretations were represented along three dimensions: information selection, relative salience, and relational organization. The relational dimension was partially operationalized through co mobilization. These representations were aggregated into fixed historical profiles and used to organize otherwise identical explanation sentences. Participants encountered four organization procedures: the OWN historical profile, the DONOR historical profile, a population profile called POP, and a language model procedure using the historical profile called PROFILE LLM. Post exposure objective comprehension was the primary outcome. OWN achieved 86.5 percent correct responses, compared with 66.0 percent for DONOR, 63.9 percent for POP, and 73.0 percent for PROFILE LLM. Mixed effects analyses showed higher comprehension with OWN than with each alternative procedure. No reliable overall condition effect was detected for recommendation acceptance. These findings indicate that historical profiles derived from prior interpretations can provide an actionable signal for personalizing explanation order while keeping explanatory content and wording fixed.
Read the full abstract
Machine Learning (Statistics)
Blind Noise Equalization for Large Data Matrices under General Variance Profiles
Low-rank approximation is a basic tool for inference from large and noisy data matrices. However, when noise variances differ across entries, a setting known as heteroskedasticity, the noise can distort the observed spectrum, producing spurious components and obscuring weak signals. To address these challenges, we develop BRINE (Blind Resolvent-based Iterative Noise Equalizer), a data-driven procedure that iteratively rescales rows and columns to equalize the average noise variance across both dimensions. Under suitable conditions, this procedure restores the Marchenko-Pastur law for the noise spectrum and can amplify signal components without prior knowledge of the signal or noise structure. A key idea underlying our approach is to equalize the diagonal entries of the data resolvent as a proxy for equalizing the unknown noise variances. At the population level, we identify a potential function that characterizes this equalization and motivates an iterative minimization procedure. We prove convergence of the iterations to the true equalizer for arbitrary positive variance profiles, with a geometric rate under mild conditions. The updates can be estimated directly from noisy data, and we establish probabilistic error bounds for the scaling factors. Simulations and experiments across diverse biological data modalities and preprocessing pipelines illustrate the effectiveness of our approach.
Read the full abstract
Signal Processing
Multi-Model Over-the-Air Federated Learning via Transceiver Beamforming and Device Grouping
This paper considers uplink communication design for training multiple independent machine learning models via wireless federated learning. To reduce communication delay, we utilize multi-antenna beamforming and propose a multi-model over-the-air (M-OTA) uplink transmission and aggregation method for simultaneous training of multiple models. Devices are formed into groups for local model training based on the model assignments per communication round and simultaneously perform OTA aggregation over a multi-access channel. Based on the M-OTA transmission and reception process, we formulate a joint optimization of transmit and receive beamforming and device grouping over $T$ communication rounds to maximize the training convergence rates of all models. We analyze the per-model optimality gap, which captures the impact of noisy aggregation and inter-model interference on the learning performance. We further show that the $T$-round joint optimization problem can be decomposed into per-round online optimization to devise solutions. Following this, we propose a device grouping strategy that explores channel correlation among devices to maximize OTA aggregation performance. Furthermore, we present a low-complexity iterative algorithm for joint transmit-receive beamforming with closed-form updates. Simulation shows that our proposed M-OTA approach substantially outperforms the conventional single-model sequential OTA approach and other commonly used beamforming schemes.
Read the full abstract
Computer Vision
Extrapolation Matters: Sparse-Only Depth Completion Beyond LiDAR Coverage
Extrapolation beyond LiDAR coverage is an important yet under-evaluated challenge of depth completion. In real camera–LiDAR systems, partial sensor overlap, occlusions, and limited field coverage often leave large image regions weakly observed or completely unobserved by LiDAR. Thus, we investigate depth completion in this beyond-coverage regime using only sparse supervision from raw single-sweep LiDAR, without dense depth labels, LiDAR accumulation, or pseudo-LiDAR generation. Our approach leverages a frozen monocular depth foundation model, lightweight per-frame alignment to sparse LiDAR geometry, and a refinement module that fuses the aligned prior with sparse LiDAR anchors to predict dense metric depth. To regularize learning under raw sparse supervision, we introduce a high-frequency normal-consistency loss defined as the image-space gradient of the per-pixel normal cosine discrepancy. This term penalizes spatial variation in scalar angular error; it does not directly match changes in neighboring normal vectors. We further introduce an evaluation protocol that explicitly separates LiDAR-supported and beyond-coverage regions, revealing extrapolative behavior that conventional depth-completion metrics can obscure. Our method achieves strong depth accuracy beyond LiDAR coverage while remaining competitive in LiDAR-covered regions, providing a practical way to deploy monocular foundation priors in multi-sensor systems.
Read the full abstract
Artificial Intelligence
Aligning, Fast and Slow: A Control-Theoretic Oversight Framework for Long-Horizon Trajectories and Maintained Deployments
AI agents can pass checks on individual tasks while retaining memory or code changes that alter their later behavior. We introduce a control-theoretic framework that couples a fast loop controlling proposed actions with a slow loop reviewing persistent changes and reassessing configurations after model or tool updates. When measurement errors and changes between checks are bounded and controllers can enforce their decisions, we derive a deployment-level bound and a rule for dividing review effort. In simulation, dividing a fixed review budget between both loops reduces mean maximum departure from intended behavior roughly twelvefold compared with review confined to individual runs. In coding stress tests with authored memories and five matched seeds, combined controls reduce mean severity on a 0-3 scale from 0.325 without gates to 0.117, resolving 33 of 120 issues versus 35 without gates; in trading stress tests, combined controls reduce mean shortfall in low-emission purchases from 0.892 without gates to 0.107. The overall framework helps developers and auditors compare and enforce oversight designs, while the agent studies test component controls rather than the full deployment-level guarantee.
Read the full abstract
Computation and Language
Guardrail Erosion in Domain Fine-Tuned Large Language Models: A Paired Audit of Knowledge-Centric Abstention in Biomedical LLMs
The number of domain fine-tuned models, particularly in the biomedical domain, is rapidly expanding. Fine-tuning promises high-performing models at a fraction of the cost of training a base model, but it can also produce unintended behavioral changes. Multiple studies have shown that safety guardrails can erode with fine-tuning, yet large-scale studies across model architectures, data mixtures, and fine-tuning methodologies remain limited. Where prior work in this space has studied value-centric refusal, this work focuses on knowledge-centric abstention. We audit 25 biomedical domain fine-tuned large language models against 11 paired base checkpoints on MedAbstain and ClinDet-Bench to evaluate the impact of fine-tuning on knowledge-centric abstention. We find that defined appropriate abstention measured by these benchmarks, already low in base models (median 8.3% on MedAbstain), falls further after fine-tuning (median 4.2%, decreasing in 21 of 25 pairs on MedAbstain), and that the magnitude and direction of change are heterogeneous across pairs. These results motivate fine-tuning methods that improve domain-specific capability while preserving appropriate caution.
Read the full abstract
Artificial Intelligence
Jacobian-Causal Memory Control for Persistent Memory in Language Model Systems
Persistent-memory systems choose not only which records are retrieved but which prior information can alter current inference. We propose Jacobian-causal memory control (JCMC), which represents each authorized memory intervention by a predicted influence signature in a shared behavioral measurement space, selects interventions against a target effect and side-effect budget, and compares predicted with observed effects before retaining or rolling back the intervention. Under explicit smoothness and bounded-estimation assumptions, we prove a quadratic local-approximation bound, characterize the target effects reachable by a set of signatures, bound selection regret under signature error, and bound non-additive interactions between simultaneous memories. The paper reports no benchmark measurements; it defines a paired counterfactual estimator and an evaluation protocol comparing JCMC with relevance, graph, and policy-filtered retrieval baselines. The central claim is narrowed or falsified if those simpler baselines match effect targeting, side-effect control, and rollback utility under equal model, memory, policy, and cost budgets.
Read the full abstract
Computation and Language
Semantic Conditional Similarity (SCS): A Formal Mathematical Framework
Semantic Textual Similarity (STS) operates on document pairs, returning a single scalar per comparison. In a multi-document setting where several sources may have contributed to one or more targets, pairwise STS cannot disentangle each source's marginal semantic contribution from their shared vocabulary. In this paper, Semantic Conditional Similarity (SCS) is introduced as a conditional form of STS that computes similarity between two documents after eliminating semantic elements shared with a third via set-theoretic operations. SCS makes four contributions: 1) it introduces directionality, partitioning pairwise relations into lateral (inter-source) and vertical (source-to-target) classes; 2) conditioning induces role-dependent asymmetry by extending the comparison domain from unordered pairs to ordered, conditioned triples of a target, a reference, and a conditioning document; 3) SCS introduces inter-document conditioning by subtracting shared content from a third document, complementing the intra-document conditioning in recent C-STS work; and 4) the set-theoretic decomposition achieves threshold-free partitioning of shared and residual content, requiring no arbitrary similarity cutoff. Although developed for the $n \to 1$ configuration, the framework extends to $1 \to n$ via vertical role reversal and to $m \to n$ via independent composition of source- and target-side conditioning. With the established $1 \to 1$ case, these span the full spectrum of document cardinality. SCS scores should be interpreted ordinally, as rankings of marginal contributions rather than absolute measurements. The broader implication is that metric space axioms commonly assumed in similarity formalizations constrain expressiveness in role-differentiated multi-document settings, and that extending the domain to conditioned triples can recover information that pairwise-symmetric comparisons suppress.
Read the full abstract
Systems and Control
Reducing Data-Center Load Transfers with Ride-Through-Aware Voltage Control
Fault-induced transfers of large data centers to backup supply can cause abrupt reductions in grid-served demand. Coordinating reactive-power support to mitigate these transfers requires consideration of facility-specific voltage–duration ride-through characteristics, particularly when support capacity is limited. This paper develops a ride-through-aware extension of quadratic voltage control that assesses voltage deviations relative to the admissible margins specified by these characteristics. We establish four design requirements that jointly express envelope-based severity, duration dependence, and preventive voltage-restoration incentives. Within a multiplicatively gated quadratic class, we derive necessary and sufficient conditions for satisfying these requirements and construct an explicit family of envelope-normalized objectives. We apply the proposed objectives to reactive-power coordination and evaluate their performance on the IEEE 14-bus, New England 39-bus, and NPCC 140-bus systems. Comparisons with Volt/VAR control, conventional nominal-voltage tracking, and recent data-center voltage-control methods demonstrate improved data-center retention under matched reactive-power support constraints. In the IEEE 14-bus study, the proposed controller reduces the mean data-center disconnection rate by 44.8% and grid-demand loss by 27.4% compared with the best-performing baseline. These results highlight the benefit of incorporating data-center ride-through requirements into reactive-power allocation.
Read the full abstract
Nuclear Theory
Features of Non-Local Quantum Correlations in Nuclei
Classical simulation of quantum many-body systems is challenged by the combined effects of entanglement and non-stabilizerness (magic). Non-local magic, $M^{\mathrm{NL}}_\alpha$, quantifies the component of non-stabilizerness that cannot be removed by local unitary transformations and provides a measure of quantum complexity beyond entanglement alone. It is bounded from below by the antiflatness of the entanglement spectrum, ${F}_A$. We calculate stabilizer Rényi entropies, non-local magic, entanglement-spectrum antiflatness, and entanglement entropy in shell-model ground states of more than 260 even-even, odd-even, and odd-odd nuclei, using a bipartition into proton and neutron sectors. We find that the previously established bound $4{F}_A\leq{M}^{\mathrm{NL}}_{\mathrm{lin}}$ holds for every nucleus studied and approaches saturation in the limit of separable states. Odd-odd nuclei exhibit markedly different patterns of entanglement and magic compared with even-even and odd-even systems, revealing distinctive signatures of proton-neutron quantum correlations and providing new insight into the complexity generated by nuclear interactions.
Read the full abstract
Machine Learning
On Lipschitz Explosion in Deep Neural Networks with Normalization: Consequences for Optimization and Adversarial Robustness
Can a network look stable at every layer and still be exponentially unstable as a whole? The Lipschitz constant of a neural network is a basic stability quantity appearing in optimization guarantees, generalization bounds, and robustness certificates. Deep networks naturally admit two such notions, input Lipschitzness, measuring sensitivity to data perturbations, and parameter Lipschitzness, measuring sensitivity to weight perturbations. We prove that, for deep networks with normalization layers, with batch normalization as the canonical example, both Lipschitz constants can grow exponentially with depth and hence with parameter dimension, even under strong per-layer norm control such as $\|W_\ell\|_2\le 1$, and already for linear activations. Thus, a mechanism widely used to stabilize training can create severe worst-case instabilities that are invisible locally, i.e., from layerwise norm bounds alone. Parameter-Lipschitz explosion turns Lipschitz-dependent nonsmooth optimization guarantees into exponential-in-dimension bounds for deep normalized networks, while input-Lipschitz explosion makes worst-case Lipschitz-based generalization and robustness certificates vacuous. The latter also yields a concrete rank-separation mechanism for adversarial vulnerability. The theory predicts that perturbations introducing new singular directions should be amplified much more strongly than perturbations at the same noise level that remain within the input's original singular subspace. Experiments on MNIST, Fashion-MNIST, and CIFAR-10 support this prediction, showing that rank-creating perturbations cause substantially sharper drops in accuracy, confidence, and margins than same-subspace perturbations.
Read the full abstract
Cryptography and Security
Gain-Dependent Entropy Characterization of a Zener Avalanche Noise TRNG via Dual NIST Standards: Toward Low-Cost Hardware Entropy for ICS Deployments
True random number generators based on Zener avalanche noise offer a low-cost entropy source for industrial control system and IoT edge devices, yet the relationship between amplifier gain, coupling capacitance, and entropy quality remains uncharacterized under current NIST standards. This paper presents a comprehensive entropy assessment of an op-amp-based Zener avalanche noise generator across five gain configurations and two coupling capacitor values, evaluated against the NIST SP 800-22 statistical test suite and the SP~800-90B entropy source assessment using 1.44 billion raw bits per configuration. We identify a three-phase autocorrelation structure, attributed to comparator saturation-recovery dynamics, consistent across all gain settings. A direct-current substitution experiment, replacing the Zener diode with a stable voltage source at the same bias point, reduces switching activity by approximately 97%, confirming that Zener avalanche noise, rather than amplifier self-excitation, is the necessary entropy source. The coupling capacitor is further shown to govern comparator overdrive and entropy quality independent of this source mechanism, with the smaller capacitance improving entropy, throughput, and bias. The combined results identify an optimal gain configuration achieving a perfect statistical test score, NIST-compliant min-entropy, and peak entropy throughput at a bill-of-materials cost under five dollars, offering practical design guidance for low-cost entropy sources in industrial and embedded deployments.
Read the full abstract
Dynamical Systems
Cantor's onion: a uniformly quasiregular map whose Julia set is a Cantor set of spheres
For integers $d,m\ge2$ with $1/d+1/m<1$, we construct a uniformly quasiregular map of $\mathbb{S}^3$ of degree $d^2+m^2$ whose Julia set is a Cantor set of concentric round spheres. To our knowledge, this is the first example of a uniformly quasiregular map in dimension three with infinitely many Fatou components. The map is a three-dimensional analogue of McMullen's maps $z^d+\lambda/z^m$, built from two Zorich-type power maps glued across a single shell by a slit construction. Its Julia set has Hausdorff dimension $2+s$, where $d^{-s}+m^{-s}=1$, and the restriction of the map to its Julia set is a skew product over the shift on two symbols and has topological entropy $\log(d^2+m^2)$. We also prove that uniformly quasiregular maps that expand uniformly near their Julia set, possibly with branching there, have no wandering domains and only finitely many periodic Fatou components, all of them attracting basins. The map constructed here belongs to this class.
Read the full abstract
Networking
StarLoc: A Practical Approach to Encoding Geographic Locators in IPv6 for Starlink
LEO satellite networks such as Starlink provide Internet connectivity across continents, oceans, and airspace. IP geolocation providers map Starlink's IPv6 prefixes to locations mainly through its self-published GeoFeed. However, users may be far away from their assigned Points of Presence (PoPs), one label covers an entire prefix, and labels and PoP mappings can change while addresses stay the same. We characterize Starlink's IPv6 addressing using 925 historical GeoFeed snapshots (2024–2026), one year of PoP mappings, and over 7.9 million observed IPv6 addresses. Our findings reveal that customer prefixes lack clear geographic structure, prefix labels lie a median of 540km from their PoPs, and relabeling a single /40 can affect the reported locations of up to 65,536 possible /56 delegations. To address these challenges, we propose StarLoc, a geographic Locator–Identifier separation scheme that embeds terminal location into IPv6 addresses using a resolution-adaptive H3 Locator while preserving existing Starlink prefixes, /56 delegations, and stable Endpoint Identifiers. To balance accuracy and privacy, StarLoc reveals only cells with at least a certain number of terminals, and operators can cap resolution in sensitive regions or withhold location entirely. Unlike topology-based schemes, StarLoc encodes physical location and lets Locators follow moving users.
Read the full abstract
Computer Vision
Federated training performs as well as centralized for 3D Latent Diffusion for Diagnosis-Conditioned Brain MRI Synthesis
Generative models of brain MRI are useful to understand brain conditions, and to enrich scarce and imbalanced neuroimaging data. However, training these models requires large, multi-site training sets, which is challenging. Federated learning (FL) allows for training deep learning models without sharing data. We show that training a 3D conditional latent model federatedly (where data is kept private at each site) performs similarly to centralized training (where all the data have been copied to a single site). We evaluate three metrics: conditional fidelity (how well generated MRIs perform in classification), diversity (how the generated images match the real data distribution, using MultiScale-SSIM), and realism (how close generated images are to real images, using Fréchet Inception Distance). Our results show that federated training is a viable route to develop multi-site generative models of brain MRI.
Read the full abstract
Distributed Computing
ReKV: Overlapping Quantized KV Transfer and Suffix Recomputation
Prefill–decode disaggregation moves a prompt's key–value (KV) cache between workers before the first token can be generated. Reducing this handoff to either cache compression or recomputation misses a useful choice: which layers to transmit, at what precision, and which layers to execute again on the decode worker. ReKV transfers a quantized prefix KV cache and one BF16 boundary state, then reconstructs a contiguous suffix while the prefix is in flight. In a model-shaped synthetic benchmark on cross-node H100s, the Q8 configuration meeting separately evaluated quality criteria reduces 8K handoff latency by $2.36\times$, $2.84\times$, and $3.65\times$ over BF16 transfer for Qwen2.5-7B, Qwen2.5-3B, and Qwen3-8B. It is also faster than the best measured transfer-only or recompute-only action at those points. A separate real-weight BF16 recomputation experiment reduces prefill-to-token-return time by $1.16$–$1.63\times$. The preferred action changes with prompt length and link bandwidth.
Read the full abstract
High Energy Astrophysical Phenomena
Bicoherence States from QPO-Accretion Flow Coupling in Black Hole X-ray Binaries
We calculate the bispectrum of a black hole X-ray binary light curve model where lognormal fluctuations associated with the inner flow are multiplicatively modulated by a quasi-periodic oscillation (QPO) and its harmonic. The bispectrum is a third-order Fourier statistic which measures the strength of nonlinear relationships between frequencies in a timeseries. When applied to black hole X-ray binaries, the bispectrum reveals nonlinear coupling between QPOs and the broadband noise. These couplings are sorted into a small set of states and X-ray binaries are observed to transition between these states as they undergo spectral transitions. We show that the model we consider is able to reproduce all of the canonical bispectral states analytically. We consider other sources of incoherent variability which dilute this model light curve and find that changes in this dilution, together with changes in the QPO amplitude, drive the transitions between the bispectral states. We also show that the larger QPO amplitudes of high-inclination sources favor the cross state and the weaker QPOs of low-inclination sources favor the hypotenuse state, which can explain the empirically observed correlation. Additionally, any dilution by equatorial winds, which are seen mainly at high inclination, would push high-inclination sources further toward the cross state.
Read the full abstract
Cryptography and Security
Conflux: Preventing Privilege Escalation with Principal Contexts and System-Level Evaluation
LLM agents combine information from multiple principals while acting externally. We formalise privilege escalation (PE) as the system-level security objective: an action may proceed only if every principal that influenced it would allow it. Model-level defences remain probabilistic, so enforcement must sit outside model judgement. We present Conflux, a principal-context defence. Its kernel, ITES, derives authority from provenance and the organisation's existing access-control system and intersects the permissions of every influencing principal. We prove a conditional authority-confinement theorem: under complete mediation, no-laundering provenance, a known non-empty decision context, fresh permission checks, and action binding, the kernel admits no action that escalates privilege with respect to any recorded influencing principal. The theorem covers the represented value-influence fragment; control influence from tool, branch, recipient, or plan-step selection remains a refinement obligation. Conflux extends the kernel with independent policy ports, argument-aware authorisation, disclosure, consent, delegation, restriction, and exposure planning. We describe two system- level evaluators. SLED is a black-box evaluator: it exhaustively explores a declared finite model of a defence up to a bound. SLED-V is a white-box evaluator: it checks a serialisable intermediate representation against security and provenance properties. The evidence is bounded and partial. The appendix reports retained AgentDojo, SLED, and SLED-V results, including the limits of those runs. No result in this paper is an unbounded deployment guarantee.
Read the full abstract
Machine Learning
MEDALLION: Manifold Embedding Distillation and Attributions for Low-dimensional Latent InterpretatiON
Dimensionality reduction is widely used to visualize high-dimensional data and guide scientific discovery, yet interpreting which original features contribute to low-dimensional representations remains challenging, particularly for black-box methods such as t-SNE and UMAP. We introduce MEDALLION (Manifold Embedding Distillation and Attributions for Low-dimensional Latent InterpretatiON), a novel framework for interpreting low-dimensional embeddings through knowledge distillation and neural feature attribution. MEDALLION distills any given embedding with a neural encoder which we explain via a broad range of existing neural feature attribution methods. Through experiments on MNIST and applications in single-cell transcriptomics, we show MEDALLION provides informative feature attributions that cast additional insights into the underlying embedding and compare its performance with existing interpretability approaches.
Read the full abstract
Solar and Stellar Astrophysics
The Complex Magnetic Anatomy of a White Dwarf Chromosphere
This study presents phase-resolved spectroscopy and spectropolarimetry of the DAHe white dwarf LP 705-64. Here, two apparently discrete chromospheric regions are observed to alternate in visibility and magnetic polarity with rotation, with field strengths of 9.5 and 22 MG based on the measured Zeeman splitting in H{\alpha}. As in the class prototype GD 356, emission line equivalent widths vary in antiphase with the broadband light curve. The 9.5 MG field is visible for 60 per cent of the stellar rotation and stays nearly constant, whereas the 22 MG field increases modestly during the remaining phases. Notably, the time-series transitions between magnetic regions are abrupt and transit-like in morphology, requiring at least one magnetic region to be highly localised and significantly smaller than the areas inferred for the dark surface spots. Both the intensity and the Stokes V/I spectra indicate a field configuration significantly more complex than an offset dipole, but higher-cadence data are necessary to distinguish between a global field and localised magnetic regions. All data are consistent with magnetic, chromospheric, and dark surface regions that are co-located and thus physically related. These findings support a model of chromospheric emission that is intrinsic and localised, consistent with predictions of dynamo-driven electrical currents that heat the upper atmosphere near magnetic poles.
Read the full abstract
Analysis of PDEs
Global regularity of the two-dimensional fractional Boussinesq equations with critical dissipation
In the present paper, our primary focus is on the Boussinesq equations with fractional velocity dissipation $\Lambda^\alpha u$ and temperature diffusion $\Lambda^\beta\theta$. On the critical balance line $\alpha+\beta=1$, establishing global regularity for weak velocity dissipation has been obstructed by the quadratic convective-buoyancy feedback, and all available results below $\alpha=\frac{10}{13}$ require a smallness assumption on the initial temperature. In this paper, we prove large-data global regularity across the extended regime $\alpha>\frac{\sqrt{57}-5}{4} \approx 0.637$. This provides a step beyond the previous barrier $\alpha=\frac{10}{13}$, unconditionally dismantling the temperature smallness assumption that has persisted in the literature for over a decade. The mathematical mechanism relies on a high-order Quadratic Normal Form transformation $H=\omega-R_{\alpha}\theta-\Bop(\theta,\theta)$ tailored to eliminate non-resonant quadratic vorticity-temperature interactions. Crucially, by uncovering the exact physical scalings of the underlying multilinear commutators and maintaining sharp fractional conjugates in the $L^m$-energy framework, we ensure that the global energy closure is governed entirely by the quadratic thermal dissipation. Once an integrability threshold $m>\frac{2}{\alpha}$ is attained, a fine Littlewood-Paley transport estimate drives a fractional regularity gain $H \in B_{r,\infty}^{3\alpha-2}$, ensuring the velocity field achieves critical drift regularity $C^{\alpha}$. By Silvestre-type drift regularization, this spontaneously generates positive Hölder continuity for the temperature, which in turn elevates the drift regularity strictly beyond the critical threshold, unlocking non-local Schauder estimates to prevent finite-time singularity formation.
Read the full abstract
Commutative Algebra
Pruning free resolutions over a complete intersection
Given a free resolution $F_\bullet$ over a polynomial ring $R$ and a system of higher homotopies on it for a regular sequence $a_1,\dots,a_c$, we can use the Eisenbud-Shamash construction to build a free resolution over $\overline{R}=R/(a_1,\dots,a_c)$. We provide a combinatorial formula for the new system of higher homotopies on a pruned free resolution $F_{\bullet}^{\mathcal{M}}$ that corresponds to a pruning of the Eisenbud-Shamash resolution.
Read the full abstract
Graphics
CrowdDirector: Runtime Orchestration of Crowd Simulation through Natural-Language Directives
Natural language offers an intuitive way to direct a virtual crowd while a simulation is running. Handwritten rules execute efficiently but cannot respond flexibly to varied instructions, whereas asking a large language model (LLM) to control the crowd at every update introduces repeated latency, cost, and the risk of unsupported actions. We present CrowdDirector, a framework that uses an LLM only when the user adds or changes an instruction. The LLM translates the request into a structured instruction, and a scene validator checks whether the entities and actions it references exist in the current Unity scene. Accepted instructions are stored and remain continuously active as the simulation is running. At each simulation tick, a graph neural network uses the current scene state and active instruction to choose an available action and target for each agent, while Unity handles movement, interactions, and animation. The network is trained offline on simulated scenarios, with LLM guidance on how agents should respond to instructions. Our evaluations show the scene validator's feasibility decisions matched human judgements more often than the LLM's own assessment, and on a behavioural benchmark CrowdDirector follows user instructions more closely than handwritten rules while avoiding repeated LLM calls. It therefore provides a practical middle ground between fixed rules and continuous LLM control.
Read the full abstract
Systems and Control
Robust Adaptive Lyapunov-Based MPC for Nonlinear Systems with Experimental Validation on Quadcopter Attitude Control
In this paper, a computationally efficient robust adaptive Lyapunov-based model predictive control (RALMPC) framework is proposed for a class of constrained nonlinear systems subject to uncertainties and external disturbances. A radial basis function neural network (RBFNN), introduced to approximate non-parametric uncertainties, is incorporated into a control Lyapunov function analysis, together with integral barrier Lyapunov functionals for state constraint enforcement. This analysis yields control and adaptation laws for parametric uncertainties, external disturbances, RBFNN weights, and worst-case approximation error, rendering the closed-loop system uniformly bounded while ensuring convergence of tracking errors to compact sets around the origin. The resulting guarantees are then embedded into the RALMPC optimal control problem through a contractive constraint, thus combining formal closed-loop stability with real-time performance optimization; at the same time, the adaptation laws allow the RALMPC prediction model to be updated online. Comparative hardware-in-the-loop simulations against state-of-the-art MPC schemes validate the efficacy and computational efficiency of the proposed framework. Ultimately, RALMPC is experimentally validated on a custom-built quadcopter platform for time-varying attitude tracking, demonstrating its practical feasibility; comparisons against alternative strategies confirm its superiority in achieving improved closed-loop performance.
Read the full abstract
Materials Science
Bichromatic light-induced electronic torques on the Néel vector in Rashba magnetic multipoles from $p$-wave magnets to $i$-wave altermagnets
Light can control magnetic order on ultrafast timescales, but for a two-dimensional magnet with an out-of-plane order parameter $\boldsymbol N\parallel\hat z$, the inverse Faraday field is collinear with $\boldsymbol N$ and gives no transverse torque. Here we show that bichromatic $(\omega,b\omega)$ light generates electronic torques perpendicular to $\boldsymbol N$ in Rashba-coupled magnetic multipoles, from $p$-wave ($n=1$) magnets to $i$-wave ($n=6$) altermagnets. Beyond high-frequency effective fields, we evaluate the exact Floquet response and the generalized torque per occupied electron, for prescribed occupations. The momentum-dependent exchange terms dominate this Fermi-sea torque and remove the leading-order zeros (amplitude-ratio blindspots) of the momentum-independent field in the channels tested. For out-of-plane $\boldsymbol N$, rotational symmetry permits a transverse torque only if $\gcd(n,b+1)=1$ for counter-rotating beams or $\gcd(n,b-1)=1$ for co-rotating beams. For in-plane $\boldsymbol N$, the Fermi sea also gives a torque component along $\hat z$ that the momentum-independent field cannot produce. We classify the active channels, with $n=1,\dots,6$ and $b=1,\dots,7$, by frequency scaling and origin (cycle-averaged versus oscillating exchange). As a control, a Rashba ferromagnet ($n=0$) has one symmetry-allowed out-of-plane channel and a weak in-plane response without a selection rule. Finite-momentum occupied-state physics is therefore essential for light-driven torques in unconventional magnets, and the exact Floquet calculation validates and extends the high-frequency description.