Machine Learning
Papers
Fresh from arXiv, typeset properly. Highlighted words trade on the market.
All of arXiv
Saturday 12 September
On-policy distillation (OPD) is increasingly applied to frontier foundation model post-training. Prior work in this area has largely focused on algorithmic advances, yet it remains unclear how much OPD depends on its training questions and, in particular, how far this dependence can be reduced. Across two representative single-teacher OPD settings, we find that training on 8 real prompts yields performance comparable to training on 17k problems, while datasets differing substantially in measured difficulty and initial distillation gap yield similar outcomes. Our analyses suggest two complementary explanations: repeated sampling could allow even a few prompts to expose substantial teacher supervision, while OPD transfers generalizable reasoning capabilities beyond dataset-specific knowledge. Building on these observations, we next propose a data-free on-policy distillation (DF-OPD) setting to investigate whether the system can supply the training questions itself, eliminating the need for external data. With 64 self-generated questions obtained without seed examples, DF-OPD yields performance comparable to full-data OPD in both single-teacher settings. This finding also holds in multi-teacher OPD: across mathematics, code, and instruction following, 1k generated questions achieve performance comparable to training on approximately 7k real post-training examples. We further explore whether OPD can operate even without explicit training questions. The experiments show that this is effective only in limited cases, where the student unexpectedly generates and answers its own questions, thereby reducing the process to an implicit form of DF-OPD. Together, these findings invite a reassessment of the role of training data in on-policy distillation. Code is available at https://github.com/Ryuki661/DF-OPD
Optimization and Control
Resilient Extremum Seeking Control for Cyber-Physical Systems Under Denial-of-Service Attacks
Extremum seeking control (ESC) relies on deliberately injected excitation to extract optimization information from measured outputs, making its networked implementation particularly vulnerable to denial-of-service (DoS) attacks. This paper develops a resilient discrete-time ESC architecture for cyber-physical systems subject to DoS attacks. By holding the most recently successfully transmitted signals during DoS intervals, the proposed mechanism yields averaged error dynamics with an exponentially contracting mode under successful communication and a neutral mode under attack. For deterministic DoS attacks, practical exponential convergence to the extremum is established under an average bound on the attack duration, with an explicit convergence rate depending on the fraction of time under attack. For probabilistic DoS attacks, almost-sure and in-probability convergence properties are established via stochastic averaging, with the attack success probability explicitly entering the convergence rate. A key finding is that a seemingly natural zero-input strategy can fundamentally compromise ESC: attacks synchronized with the dither generate an $\mathcal{O}(1/a)$ bias in the averaged dynamics, where $a$ is the dither amplitude, and can displace the equilibrium from the true optimizer. Thus, under the proposed hold-based architecture, increasingly severe DoS attacks primarily slow the optimization process rather than destroy its stability, provided that communication is not permanently blocked. Numerical simulations illustrate the theoretical guarantees and the failure mechanism induced by dither-synchronized attacks.
Read the full abstract
Strongly Correlated Electrons
Haldane-Holstein model at fractional filling: Route to bosonic fractional Chern insulator and quantum anomalous Hall crystal
Electron-phonon coupling is generally expected to suppress band topology, driving a topological band insulator into a trivial phase. Here we show that, using the prototypical Haldane-Holstein model at partial filling of a topological band, strong electron-phonon coupling can instead stabilize topological phases of matter. The electron-phonon coupling plays a double role: it generates the longer-range interactions that correlate the carriers and, in the attractive channel, also binds them into bosonic pairs. For spinful electrons, it pairs opposite spins into charge-$2e$ bipolarons that form a bosonic fractional Chern insulator (FCI) at filling $\nu=1/2$ with charge-$e$ semionic excitations. Tuning the band topology and the coupling strength maps out a rich phase diagram containing this bosonic FCI, superconductors condensed at the $M$ and $K$ points, and several charge-ordered solids. For spin-polarized electrons, strong coupling instead stabilizes a $C=2$ quantum anomalous Hall crystal (QAHC) at $\nu=1/3$. Electron-phonon coupling thus emerges as a route to, rather than an obstruction against, topology in partially filled Chern bands.
Read the full abstract
Group Theory
Conjugator length in nilpotent groups
For every rational number $\alpha\geq 2$, we construct a 2-step nilpotent group with conjugator length function $\operatorname{CL}(n)\simeq n^\alpha$. We deduce that the nilpotent conjugator length spectrum is dense in $\{0\}\cup[2,\infty)$, even after restricting to groups of nilpotency class at most $2$. Moreover, we show that, despite being a commensurability invariant of nilpotent groups, conjugator length is not a quasi-isometry invariant: for every $m \geq 1$, the cocompact lattices in $\mathsf H(\mathbb R)^m$ realize exactly the growth types $n^2,\ldots,n^{m+1}$. Finally, for every real cubic algebraic number $\theta$, we construct a 2-step nilpotent group such that $n^3 \preceq \operatorname{CL}(n) \preceq n^{3+\varepsilon}$ for every $\varepsilon > 0$, with $n^3 \prec \operatorname{CL}(n)$ if and only if at least one real conjugate of $\theta$ has unbounded partial quotients. Consequently, determining the conjugator length function for arbitrary 2-step nilpotent groups is at least as hard as settling the bounded-partial-quotient problem for real cubic algebraic numbers with two nonreal conjugates.
Read the full abstract
Computer Vision
MorphoStyle: Motion Style Transfer with Morphology Control
Human motion may be viewed as a combination of action content, style, and body morphology. Existing motion style transfer methods transfer a reference style onto a content motion while assuming a canonical body, whereas shape-aware motion generators adapt motion to a target shape without explicit style control. This separation of motion style and shape (morphology) makes it difficult to generate stylized motions for non-canonical bodies; naively combining a style transfer module with a shape-aware generator often leaks action content from the style reference and disrupts shape-consistent kinematics. We present MorphoStyle, a framework for shape-aware motion style transfer that is built on a shape-conditioned Finite-Scalar-Quantization Variational Auto-Encoder (FSQ-VAE). The key contribution is to pose the desired style transfer as modular latent disentanglement comprising: (i) a contrastive style encoder that extracts content-decoupled style embeddings; (ii) a text-guided style-routing mechanism that locates style-relevant joints in a text-motion feature space; and (iii) a manifold preserving style modulator that injects discriminative style embeddings in content features as a temporally-gated low-rank offset. Extensive experiments on benchmark datasets demonstrate that MorphoStyle outperforms competing baselines in terms of both shape control and motion style transfer, while simultaneously providing quantitative shape control. For more details, please see project website: https://github.com/funkdub/MorphoStyle.
Read the full abstract
Number Theory
Golden-ratio growth of Conway's subprime closure
Let $s(m)$ be the Conway subprime function define the binary operation on the natural numbers $x \circ y= s(x + y)$, and denote by $C_n$, $n \ge 0$, the sequence of subsets of natural numbers defined by $C_0 = \{1\}$, and $C_{n+1} = C_n \cup (C_n \circ C_n)$. We prove the conjecture by Caragiu, Vicol and Zaki that $$\lim_{n\to \infty} \frac{ | C_{n+1}| }{ | C_n |}= \frac{1+\sqrt{5}}{2}.$$ The underlying mathematical proof in this paper was constructed with some algorithmic assistance from GPT-6 Astra and its correctness has been formally verified using the Lean 4 proof assistant.
Read the full abstract
Machine Learning
Entropy-Punctured Bloom Filters for Memory-Efficient Machine Learning
Memory-efficient feature representations are increasingly important in machine learning settings where storage, transmission cost, bandwidth, or privacy constraints limit access to raw data. Bloom Filter (BF) encodings provide compact probabilistic representations of engineered features, but their behavior under structural compression and their applicability to regression tasks remain underexplored. In this work, we propose entropy-punctured Bloom Filters, a memory-aware encoding strategy that removes low-variability bit positions identified using empirical entropy. Starting from fixed-length BF encodings of quantized features, the proposed approach produces reduced representations that preserve predictive structure while improving predictive efficiency relative to encoded representation size. We evaluate the approach on diverse regression datasets, comparing raw features, Principal Component Analysis (PCA), Random Projection (RP), and Bloom Filter variants under leakage-free evaluation protocols and approximately matched representation sizes. Performance is assessed using ridge regression, XGBoost, and neural networks, with predictive efficiency measured as R2 relative to encoded representation size per sample. Results show that Bloom Filter encodings remain competitive with classical compressed representations while achieving substantial storage savings. Entropy-based puncturing further reduces representation size with minimal loss in predictive fidelity, yielding improved predictive efficiency. These findings demonstrate that entropy-punctured Bloom Filters provide an effective representation-level compression approach for memory-constrained machine learning.
Read the full abstract
Optimization and Control
The Global Topology of Orthogonally Decomposable Tensor Landscapes
The homogeneous form associated with a symmetric tensor, restricted to the sphere, is the objective of the best rank-one approximation problem and, with random coefficients, the energy of a mean-field spin glass. Its critical points have been studied extensively, but the global organization of the landscape they form is far less understood. We study this global structure for positive orthogonally decomposable tensors. We determine the persistent homology of the sublevel and superlevel filtrations in closed form, for every homological dimension $q$, every ambient dimension $D$ and every tensor order $k$, via a recurrence in the ambient dimension. Taking the coefficients to be random, we further prove laws of large numbers for the resulting persistence diagrams at dimension $0$ and $D-2$, giving an exact description of the typical global topology of a spin-glass-like energy landscape. We illustrate our results with numerical computations and simulations.
Read the full abstract
Computer Vision
Bi-Level Routing and Sparse Spatial Attention based Multi-View BEV 3D Object Detection for Autonomous Driving
Bird's Eye View (BEV)-based multi-view 3D object detection suffers from challenges of computational complexity, multi-scale feature extraction, and efficiency of dense 2D-to-BEV view transformation. To address these problems, this paper proposes an improved BEV 3D object detection algorithm Sparse-BEVNet. Firstly, a Bi-Level Routing Attention (BRA) mechanism is introduced into the image feature extraction network to reduce the computational burden of the backbone. Second, Cascaded Group Attention (CGA) is employed in the feature fusion module, which enhances deep interaction across features of different hierarchical levels without introducing additional computational overhead. Furthermore, a Sparse Spatial Cross-Attention mechanism is adopted to replace the conventional dense view projection pipeline. Experimental results on the public nuScenes dataset demonstrate that the proposed method achieves a mean Average Precision (mAP) of 45.2% and a nuScenes Detection Score (NDS) of 54.5%, corresponding to 3.6% and 2.8% improvements relative to the baseline model, respectively.
Read the full abstract
Machine Learning
ZAPS: Zero-Cost Active Proxy Search for Neural Architecture Search
Neural Architecture Search (NAS) automates network design, but evaluating a single candidate requires training it to convergence, making exhaustive search intractable. Zero-cost proxies estimate architecture quality at initialization in seconds, yet a single proxy is noisy, and combining several does not straightforwardly help: proxies are strongly correlated, so naive aggregation compounds their shared errors instead of averaging them out. Existing methods exploit either proxy signals or architectural topology - never both within a single active-learning framework. We introduce ZAPS (Zero-cost Active Proxy Search), a four-stage pipeline that closes this gap. ZAPS (i) selects a compact, non-redundant proxy subset offline via ProxyFit, a greedy anti-redundancy criterion; (ii) seeds the search with a hybrid K-means strategy that balances exploitation and exploration; (iii) re-selects proxies at every iteration by a bootstrapped vote as the labeled set grows; and (iv) ranks candidates with an XGBoost ensemble trained jointly on proxy ranks and one-hot topological encodings, queried through an Upper Confidence Bound (UCB) acquisition function. On NAS-Bench-201 under a budget of B=200 evaluations, ZAPS recovers 52.3% of the true top-100 architectures on CIFAR-10 and 65.8% on CIFAR-100, ahead of every baseline we consider - Random Search, Local Search, REA, BANANAS and TPE - and, on CIFAR-10, with less than half the run-to-run standard deviation of the strongest of them. The advantage is largest where evaluations are scarce: on NAS-Bench-201 it narrows as the budget grows, whereas on the harder NAS-Bench-101, which no method comes close to saturating, it widens instead. All methods are scored by a single criterion: how much of the true top-100 lies among the architectures they actually evaluated.
Read the full abstract
Robotics
DreamSat-Bench: Development and Initial Testing of a Testbed for AI-Based Pose Estimation from 3D Reconstruction
This paper presents the development and initial testing of DreamSat-Bench, a modular rendezvous and proximity operation testbed designed to benchmark AI-based relative navigation techniques. By integrating a software- and hardware-in-the-loop robotic pipeline, the platform enables a seamless transition from digital simulation to physical reality. DreamSat-Bench unifies state-of-the-art robotic learning tools such as MuJoCo, Isaac Lab, and LeRobot into a single benchmarking platform, utilizing robotic arms to trace 3D trajectories. The platform allows for extensive customization of orbital environments and lighting to evaluate the simulation-to-reality gap. We demonstrate the testbed's utility by evaluating an end-to-end vision-based navigation pipeline that pairs DreamSat, a generative AI framework for single-view 3D reconstruction, with FoundationPose for zero-shot 6-DoF tracking of unseen spacecraft. Initial testing explores mission-representative orbital segments, including fixed-point station-keeping and fly-around characterization. Through a series of parametric studies, we quantify the impacts of reconstruction latency, mesh resolution, orbital range, and illumination geometry on pose estimation accuracy. Finally, a preliminary hardware-in-the-loop campaign qualitatively validates the physical deployment of the pipeline, identifying target symmetry and accumulated tracking drift as critical factors for robust navigation. DreamSat-Bench provides a rigorous framework for maturing autonomous navigation with unprepared space assets in the absence of prior geometric models.
Read the full abstract
Signal Processing
Inverse Maxwell-Based Wall-Aware OFDM-ISAC for Slow-Moving Target Sensing
In this paper, we study integrated sensing and communication (ISAC) for short-range indoor orthogonal frequency-division multiplexing (OFDM) systems in which the sensing path crosses a building wall. The target is a slow-moving user equipment behind the wall, and its echo is embedded in the wall reflection and static indoor clutter. Because the wall adds excess propagation length, attenuation, and internal reflections, a conventional delay transform reports an apparent range. We therefore formulate physical-range estimation as an inverse Maxwell problem for the known wall: the range operator is the distorted-Born Jacobian of the discretized one-dimensional Helmholtz equation, assembled offline from the calibrated wall and applied online as a per-snapshot range inversion, so that the range-Doppler map is indexed by physical rather than apparent range. We then prove that this operator carries a structural limitation: when the two-way wall factor has constant magnitude and linear phase, the wall-aware range image is identically the free-space image on a translated grid, for every ridge level, taper, and noise realization. A free-space branch translated by the same excess length is therefore a required comparison. Sensing-only numerical results for a representative layered wall, with every branch on one range grid and a common post-FFT CSI model, show that the proposed inverse removes the range bias that wall-unaware processing cannot, and that a scalar-corrected free-space inverse tracks it closely, the two differing only marginally in empirical range RMSE and detection probability across the tested sweep.
Read the full abstract
Formal Languages and Automata
Non-Terminal Complexity of Simple Semi-Conditional Grammars
We study the complexity of simple semi-conditional grammars (SSCGs) in terms of the number of their terminals and non-terminals. We show that SSCGs with three non-terminals can generate all RE languages; two non-terminals suffice for linear languages; and one suffices for unary regular languages. We also determine both upper bounds and fundamental limitations of SSCGs: while certain one-non-terminal SSCGs can generate non-context-free languages, every language generated by a one-non-terminal SSCG is in non-deterministic linear space. We also prove that there exists a regular language of alphabet size at least 3 which cannot be generated by any one-non-terminal SSCG. Finally, we prove that membership testing for SSCGs of two non-terminals is already NP-hard.
Read the full abstract
Machine Learning
A Machine Learning Framework for Fault Detection, Isolation, and Severity Prediction of Autonomous VTOL Aircraft
Fault detection in autonomous VTOL aircraft is critical because even minor component degradations can rapidly destabilize multirotor vehicles operating in complex, safety-critical environments, motivating robust fault detection and estimation strategies capable of identifying early signs of rotor damage; however, real-flight fault detection remains challenging due to sensor noise, environmental disturbances, and the nonlinear aerodynamics of multirotor platforms. This study proposes a comprehensive machine-learning framework for rotor fault detection, isolation, and severity prediction using real flight data. A convolutional neural network (CNN) architecture is developed to learn spatio-temporal patterns from multivariate flight dynamics, enabling direct inference of both the faulted rotor and its damage level. The framework is first validated using simulated data generated by a data-generative model, and experimental validation is then performed on a hexacopter by introducing controlled blade-tip breakage. The trained model achieves rotor-wise fault classification accuracies above 99% and severity estimation accuracy of 96% within a 1% tolerance in experimental data, demonstrating strong generalization and supporting real-time health monitoring for autonomous VTOL systems.
Read the full abstract
Materials Science
Ultrafast Magnetization Induced from Raman-Active Axial Chiral Phonons
Axial chiral phonons, which carry angular momentum and exhibit unusually large magnetic moments, provide a new degree of freedom for controlling the time-reversal symmetry and magnetic properties of quantum materials. While infrared-active axial phonons are efficiently excited to large amplitudes by circularly polarized terahertz pulses, their application may be constrained by short lifetimes, small penetration depths, low spatial resolution, and limited availability of optical sources. Here, we report that an axial Raman-active phonon mode in CeF3 exhibits a significant magnetic moment and an exceedingly long lifetime that more closely matches the paramagnetic spin dynamics. The axial phonon population is resonantly driven by a near-infrared laser pulse with rotating linear polarization, known as an optical centrifuge, a coherent control scheme never applied to solids before. The phonon-driven magnetization observed by time-resolved Faraday rotation scales quadratically with incident power and rapidly decreases at high temperatures, consistent with phonon inverse Faraday effect from many-body spin-phonon coupling. Our findings open a new avenue for using shaped laser pulses from widely accessible light sources to manipulate coherent axial chiral phonons and ultrafast spintronics.
Read the full abstract
Computation and Language
A Multi-Stage Agentic Framework for Effective Counter-Narrative Generation and Refinement
The rapid diffusion of hate speech and misinformation on social networks challenges democratic societies, since direct suppression efforts may deepen polarization, fuel public distrusts, and strengthen extremist narratives. LLM-driven counter-narratives (CNs) offer a promising way to reduce those risks, yet their effectiveness depends on rhetorical and stylistic choices that remain poorly understood. We present a multi-stage agent-based framework for generating, refining, and evaluating CNs, applied to pro-Russian hate and misinformation narratives on the war with Ukraine and adaptable to other domains. A pilot experiment with human evaluators identifies effective technique style pairings, such as repetition with emotional framing enhancing persuasiveness. Building on these insights, we introduce a multi-agent refinement process that iteratively improves CNs for persuasiveness, emotional engagement, and shareability. After human validation confirmed improvement, an automated safety analysis shows that our refined CNs match or improve on expert-written counterspeech. A simulated experiment then shows that they reduce the perceived strength of pro-Russian narratives and consistently outperform a vanilla LLM baseline, highlighting a pathway toward scalable, narrative-specific interventions against hate speech and misinformation. Code and data accompanying this work are publicly available at https://github.com/carmelkron/inlg2026-counter-narratives.
Read the full abstract
Signal Processing
Fourth-Order Co-Channel Interference-Aware OFDM-ISAC Sensing in Cluttered Environments
In this paper, we develop a sensing-centric receiver for monostatic orthogonal frequency-division multiplexing integrated sensing and communication (OFDM-ISAC) in the simultaneous presence of passive clutter and a non-cooperative co-channel OFDM transmitter. We show that removing the known ISAC symbols maps target and clutter echoes to deterministic delay-Doppler components, while the non-cooperative waveform remains a random, non-Gaussian, generically full-rank residual. Guided by this distinction, we combine moving-target indication for zero-Doppler clutter with a centered diagonal fourth-order statistic for interference identification. To obtain an unbiased fourth-order estimate from phase-aligned proper snapshots, we derive additive and multiplicative finite-sample bias corrections and pair the corrected statistic with a second-order moment to separate interference from Gaussian noise. Fourth order provides identification and presence testing, whereas second order provides the diagonal disturbance-power proxy used for weighting. We further introduce diagonally loaded weights that enforce a range-Doppler aperture-efficiency floor. Simulations show that the interference estimate approaches its theoretical profile as coherent support grows and remains insensitive to deterministic echoes, that the presence test resolves interference well below the level at which weighting matters, and that loaded processing improves target detection while controlling point-spread-function degradation.
Read the full abstract
Representation Theory
The Gaudin model and the super Bethe ansatz for unitarizable modules over the general linear Lie (super)algebra
Let $\mathcal{A}^{\bf z}_{\mathfrak{g}}$ be the Gaudin algebra for the general linear Lie (super)algebra $\mathfrak{g}$ with respect to a sequence ${\bf z} \in \mathbb{C}^\ell$ of pairwise distinct complex numbers, and let $\mathbb{L}$ be any $\ell$-fold tensor product of (infinite-dimensional) unitarizable highest weight $\mathfrak{g}$-modules. Fix a singular weight $\xi$ of $\mathbb{L}$. We show that the singular weight space $\mathbb{L}^\text{sing}_\xi$ is a cyclic $\mathcal{A}^{\bf z}_{\mathfrak{g}}$-module, and that the Gaudin algebra $(\mathcal{A}^{\bf z}_{\mathfrak{g}})_{\mathbb{L}^\text{sing}_\xi}$ for $\mathbb{L}^\text{sing}_\xi$ is a Frobenius algebra. We also show that $(\mathcal{A}^{\bf z}_{\mathfrak{g}})_{\mathbb{L}^\text{sing}_\xi}$ is diagonalizable with a simple spectrum for a generic ${\bf z}$. Furthermore, we establish a super version of the Bethe ansatz for $\mathcal{A}^{\bf z}_{\mathfrak{g}}$ on $\mathbb{L}^\text{sing}_\xi$.
Read the full abstract
Complex Variables
Hard Lefschetz theorem for logarithmic sheaves twisted by pseudo-effective line bundles
Machine Learning
Inherited Heads: Audio language models track speakers with their text backbone's attention, and an attention-mass ranking retrieves a different set
Asked to describe what one of six speakers in a recording talks about, audio language models describe the right one on 6 to 16% of trials, below the 16.7% a guess would give. Adding a fixed bias to the attention logits of a hundred heads, under a tenth of the model's and with no training, redirects the description to whichever speaker we choose, on 90.7% to 99.0% of trials. Those heads are largely not specific to audio. Rank the text-only language model an audio model was built from, or a released model of the same family, on a written version of the task, take its top hundred heads, and carry them over unchanged: they redirect the audio model on 80.8% to 95.0% of trials, with nothing about audio entering the selection. The audio and text head sets share 66 to 74 of 100 where chance would give about 20, and the shared part alone reproduces almost all of the steering. What that does not show is that sharing is what makes the heads work: an equal-sized draw from the same discovered hundred does nearly as well, and none of our three models separates the two explanations. A second finding concerns how such heads are found. Ranking heads by how much attention they place on the segment asked about, as an established score does, or by how much of their attention moves with the question, as a per-head normalised variant does, gives top hundreds that share 69, 37 and 4 heads across our three models. In Ultravox, where they share 4, the established score's heads leave output the judge cannot place on any segment on 69.7% of trials, against 40.0% with no intervention and 1.0% for the normalised variant. That is one arm of six; on the other five the established score steers above a random draw.
Read the full abstract
Robotics
GIFT: Glove-Inferred Force Transfer: Force-Aware Human-to-Robot Skill Transfer from a Wearable Sensing Glove to a Robot Hand Without Tactile Sensors
Human-to-robot skill transfer from sensing gloves has so far relied on shared hardware: the same tactile glove worn by the demonstrator and the robot, or a learned alignment between two tactile sensors. We present GIFT (Glove-Inferred Force Transfer), a pipeline in which the interface between human and robot is a physical unit rather than a shared sensor: fingertip force is measured in newtons on the human side and estimated in newtons on the robot side. A wearable glove records finger flexion, calibrated fingertip force, and wrist orientation, while a head-mounted camera records the demonstration; no robot is present. At deployment, the robot estimates force from actuator-current residuals relative to a free-space baseline, through a calibrated mapping to newtons, so any position-controlled hand that reports motor current can serve as the deployment platform. The policy uses a glove-space state and predicts finger-position targets; the robot enters only through two calibrated adapters, a retargeting decoder and a force estimator. We evaluate GIFT on a cup grasp-and-hold task with two action-chunking policies trained on the same demonstrations, with fingertip-force inputs retained in one and zeroed in the other. In a 50-rollout evaluation with sample size and metrics fixed before scoring, both policies succeeded in all 25 rollouts. The median of the per-rollout hold-phase grip-force estimates was 53% lower with force inputs: 1.20 N versus 2.55 N (one-sided Mann-Whitney U, p<0.0001). In an observation ablation, a vision-only policy achieved 0/15 grasps, policies given hand-command state acquired the grasp, and the force inputs determined how hard the policy held. A force channel measured on the human hand thus transfers to a robot hand with no tactile hardware, through a retargeting map from five glove channels to seven robot actuators, with no sensor shared between the two.
Read the full abstract
Mathematical Physics
Exact admissibility radii for the lattice topological charge of quantum spin textures
How much can a magnetic field change before a quantum spin texture loses its skyrmion charge? For a triangulated spin-$s$ texture on $N$ quantum sites, we determine the smallest change in its local moments that makes the reconstructed topological charge ill-defined, allowing unequal error bounds and fixed boundary moments. Combined with spectral perturbation theory, this geometric threshold yields explicit magnetic-field intervals that preserve the ground-state charge, scaling as $1/N$ when the gap and geometric margin have positive limits. We apply the bounds to finite spin-$1/2$ systems with exchange and Dzyaloshinskii-Moriya interactions. In a rigorously verified seven-spin example the charge changes at a single certified field, while the excitation gap and all local spin polarizations remain bounded away from zero; there the geometric margin vanishes linearly.
Read the full abstract
Machine Learning
CyFM: Cylindrical Optimal Transport for Few-Step Complex-Valued Flow Matching
Complex-valued signals like MRI and audio spectrograms are typically modelled as flat two-channel Euclidean data. The inherited Euclidean metric $dA^2 + A^2 d\theta^2$ vanishes at the origin, leaving phase unpenalised exactly where the signal is weakest. We replace it with the decoupled product metric $dA^2 + d\theta^2$ on the cylindrical closure $[0, \infty) \times S^1$, which stays non-degenerate at $A = 0$. We measure what this substitution costs and buys. Exact analytical bridges across synthetic fields, fastMRI knee data, and LibriSpeech spectrograms show Cartesian paths induce a heavy-tailed angular velocity distribution (Pareto index $\approx 1$). Under independent coupling, 43%-49% of signal energy falls on paths turning faster than $\pi$ rad per unit time. Cylindrical paths never reach this speed. We formulate Cylindrical Flow Matching (CyFM) to strictly bound the angular regression target, coupling noise and data via exact minibatch Optimal Transport jointly over whole fields. This coupling reduces few-step generation error by 3%-60%. CyFM achieves lower generative error than the best Cartesian baseline at every step up to $k = 8$ on synthetic fields and speech spectrograms, with all seeds separated. On knee MRI, the single-step advantage is 1.8x. At convergence ($k = 100$), the two geometries show no significant difference. Finally, a prior-only control exposes the cost of flat parametrisation: on synthetic fields, a single Cartesian Euler step performs worse than the unintegrated noise prior (0.376 vs. 0.150).
Read the full abstract
Statistical Methodology
Finite-Sample FDR Control for Greedy Aggregation over Networks
Distributed multiple testing asks $N$ sites, each holding p-values for its own hypotheses, to control the false discovery rate (FDR) of the discoveries made across the whole network while communicating only a small number of bits. Greedy interval aggregation (Pournaderi and Xiang, IEEE TSIPN, 2023) meets the communication budget but controls FDR only asymptotically, and its FDR can exceed the target at finite sample sizes; the same interval counts both select the rejection regions and calibrate the stopping rule, which biases the selected interval densities upward. We propose budgeted BONuS-GA: each site mixes synthetic uniform p-values into its data, the center ranks candidate p-value intervals across sites using the pooled counts, and each site's false discoveries are estimated from its own synthetic counts under a per-site share of the level. The procedure needs no knowledge of the null proportions, uses every p-value for both selection and inference, and controls $FDR\le\alpha$ at every sample size for any fixed assignment of hypotheses to sites, assuming only that the null p-values are independent uniforms, independent of the non-null p-values. It keeps the original $O(\sqrt m\log m)$ communication, $m$ being the total number of p-values, when $N=O(\sqrt m)$. We also give a sample-splitting alternative, cross-fit greedy aggregation, with finite-sample control under a random-site model, and we quantify the selection bias behind the original procedure's failure. In a monitoring-network simulation the budgeted procedure retains most of the power of its heuristic counterpart at moderate-to-large site sizes.
Read the full abstract
Algebraic Geometry
Discriminants of Hermitian forms, maximal degenerations and Kontsevich's tropical approach to the Hodge conjecture
Motivated by Markman's work and by Kontsevich's tropical approach to disproving the Hodge conjecture, this note explains why families of Weil type abelian 2n-folds only have maximal unipotent degeneration when the discriminant is $(-1)^n$. Along the way, it explains what the discriminant is and collects various facts about Weil type abelian varieties.
Read the full abstract
Computation and Language
Towards Evolving Context Parameterization for Large Language Models
Context parameterization enables large language models (LLMs) to internalize contexts into reusable model parameters, avoiding repeated processing across subsequent queries. However, existing methods typically assume static contexts and lack explicit mechanisms for distinguishing validity states under continual updates. To study this real-world scenario, we formalized the Memory Updating with Sequential Evolution (MUSE) task and constructed MUSE-bench to evaluate update incorporation and unaffected-information preservation. The resulting challenge requires preserving the global state while adjusting the contribution of memory evidence. Motivated by this, we proposed PLUME, a training-free method that constructs a global update representation, activates memory evidence to form a local parameter view, and adaptively integrates their predictions during decoding. Comprehensive evaluation on MUSE-bench demonstrated PLUME's effectiveness in sequential evolution settings, yielding relative improvements of 29.9% in average ROUGE-L Recall and 54.9% in LLM-as-a-Judge. Our codes are available at: https://github.com/xiaobingshi-LLM/PLUME.
Read the full abstract
Number Theory
On Decomposition of Drinfeld cusp forms of level $t$
In this article, we first prove that the Hecke operator $T_t$ has no eigenform in $\Sla$ with eigenvalue $-t^{k/2}$, when the characteristic of the base field is odd. Furthermore, if $\dim \Slt$ is even, we show that $T_t$ has no eigenform in $\Sla$ with eigenvalue $t^{k/2}$. As a consequence, we prove that the direct sum decomposition $\Slt=\Sold \oplus \Snew$ holds when $\dim \Slt$ is even. This proves the conjecture \cite[Conjecture 1.1(3)]{BV19a} of Bandini and Valentino for an infinite family of cusp forms. In particular, for any weight $k$, there exists at least one type $m$ (there are only two possible non-trivial values of $m$) for which the conjecture \cite[Conjecture 1.1(3)]{BV19a} is true.
Read the full abstract
Disordered Systems and Neural Networks
OFC-like Behavior in Experimental Granular Piles
Scale-invariant avalanche dynamics are commonly associated with criticality and robust, universal size exponents. Dissipation is generally expected to drive the system away from the critical point, progressively suppressing large events. The Olami–Feder–Christensen (OFC) model challenges this picture: in its non-conservative regime, the avalanche-size exponent is non-universal and can exceed the mean-field value $\tau=3/2$, while system-spanning events persist even at large dissipation. Here, we show that these apparently anomalous features are also observed experimentally in a two-dimensional granular system displaying scale-invariant avalanche dynamics. By increasing interparticle friction, and therefore dissipation, the avalanche-size exponent increases from $\tau=1.58$ to $\tau=1.83$, while the upper cutoff remains proportional to the system size. We further identify similarities between experiment and the OFC model in their memory effects, local dynamics, and the emergence of better-than-random predictability of large events. The latter indicates that the system does not remain permanently critical, but instead evolves through configurations with different propensities to generate extreme events. Our results suggest that OFC-like dynamics are not merely an anomalous feature of a particular model, but may provide a relevant framework for understanding scale-invariant dynamics with $\tau>3/2$ in real dissipative systems.
Read the full abstract
High Energy Physics (Theory)
Modular flow and the area operator
In the context of the AdS/CFT correspondence, the Jafferis-Lewkowycz-Maldacena-Suh (JLMS) formula relates the boundary modular Hamiltonian to the HRT area operator and the corresponding bulk modular Hamiltonian. In standard discussions of a holographic code built from perturbative excitations on a fixed classical background, the $A/4G$ term dominates over the bulk modular Hamiltonian in the limit $G \to 0$. However, this setup is insufficient to describe the action of boundary modular flows that change the classical background. In contrast, our recent discussion of `large' codes provides a useful framework to obtain a JLMS formula that relates bulk and boundary modular flows acting on appropriate states in the code. Here, one sees that the bulk modular Hamiltonian is not in fact subdominant when the modular flow defined by one state acts on another state peaked around a sufficiently different classical background. Instead, the bulk modular Hamiltonian can be of order $1/G$ and can dominate over the $A/4G$ term. Nevertheless, we show that if the two states have the same classical background except for data conjugate to the HRT area then, under mild assumptions and over a large range of flow parameters, the modular-flowed state is described by a classical background given by the HRT-area flow alone.
Read the full abstract
Number Theory
Twisted Lubin-Tate Big Witt Vectors and Fleck-Sun-Wan Congruences
We construct a big Witt theory associated with coefficient-multiplicative normalized $\sigma$-twisted Lubin-Tate formal groups. The resulting $F$-big Witt ring extends Hazewinkel's ramified $q$-typical Witt theory and recovers the usual big Witt ring in the multiplicative case. We also introduce Fleck operators defined from iterated Coleman traces and prove Lubin-Tate analogues of the Fleck-Sun-Wan congruences for both positive and negative powers. By relating the Lubin-Tate Coleman theory to the $F$-big Witt theory, we obtain a Coleman $F$-norm whose integrality yields congruences among the corresponding Fleck coefficients.