Cryptography and Security
Harvest Season for SLUB: From io_uring vulnerability to Novel Sheaf-Based Exploitation Techniques
Hao-Yu Yang and Yu-Ting Lin
The Linux kernel's push for higher I/O performance and more efficient
memory management has introduced new mechanisms that, while improving performance, also open new attack surfaces. This research examines two of them together: the io_uring subsystem and the sheaf/barn caching mechanism added to the SLUB allocator in Linux 6.18. In this research, two previously unknown vulnerabilities in io_uring are presented, and one is developed into a complete local privilege escalation chain under a hardened kernel configuration. Building this chain revealed that the sheaf/barn mechanism changes long-standing assumptions behind established exploitation techniques such as cross-cache attack, and that its design also weakens existing SLUB freelist protections. Both observations are analyzed and turned into working primitives. Building on this analysis, three novel sheaf-based exploitation techniques are proposed. Among them, an RCU-sheaf cross-cache technique removes the traditional dependence on the buddy system for moving objects across caches, giving more flexible and reliable control over object migration between cache pools. Together, these results characterize the sheaf/barn layer as a new and largely unexplored attack surface in Linux kernel exploitation.
Read the full abstract ▾
Astrophysics of Galaxies
Fooled by mass? Recovering the $\Gamma-\lambda_{Edd}$ relation of quasars from their X-ray variability
Matilde Signorini, Andrea Sacchi, Bartolomeo Trefoloni et al. (+7 more)
We measure the relation between the X-ray
photon index $\Gamma$ and the Eddington ratio $\lambda_{Edd}$ in unobscured type-1
quasars, using as parent sample the cross-match of the SDSS DR16Q catalogue with the 4XMM-DR14 serendipitous catalogue. After cleaning the sample of potential observational biases, we obtain 7835 objects, one order of magnitude larger than most previous analyses in the literature. A single regression of $\Gamma$ against $\log\lambda_{Edd}$, with single-epoch virial black-hole masses, returns slopes that depend on the mass indicator (H$\beta$, MgII, CIV) and on the bolometric correction, and that are generally flat. We interpret this as regression dilution: the systematic uncertainty on virial masses is much larger than the intrinsic spread of $\lambda_{Edd}$ in SDSS
quasars. We therefore introduce a different approach, which removes the black-hole mass from the problem. For
quasars observed by XMM-Newton at multiple epochs, the mass of each object is fixed, and changes in its X-ray luminosity trace changes in its
accretion state. We fit the $\Gamma$-$\log(L_{2keV})$ relation of each object (for a sample of N=893) and combine the per-object posteriors. This recovers the mean response of the population $a_X = 0.35\pm0.03$, which steepens to $\approx0.43$ when only epochs with precisely measured
photon indices are used. Translated to the Eddington ratio, this corresponds to $a_\lambda\approx0.20$-0.25, depending on the choice of bolometric correction and on the LX-LUV slope. The $\Gamma$–$\lambda_{Edd}$ relation therefore exists, but it can be uncovered only by exploiting the variability of individual sources, which removes the black-hole mass from the problem. Moreover, its shallowness implies that the differences in Eddington ratio, as currently measured, cannot by themselves account for the variety of
photon indices observed among unobscured
quasars.
Read the full abstract ▾
All authors Matilde Signorini, Andrea Sacchi, Bartolomeo Trefoloni, Matteo Guainazzi, Elisabeta Lusso, Emanuele Nardini, Chiara Niccolai, Federica Ricci, Anastasia Shlentsova, Mateusz Rałowski
Complex Variables
Affine balancing of complex dilatations: affine invariants and optimal distortion of harmonic quasiconformal mappings
Zhi-Gang Wang and Deguang Zhong
For orientation-preserving planar harmonic mappings, postcomposition by real-affine mappings induces
automorphisms on the space of complex dilatations. Motivated by this fact and the lack of natural affine-invariant
geometric quantities in classical distortion theory, we introduce the affine circumradius and affine diameter associated with the image of the complex dilatation. The exponentiated affine circumradius gives a characterization of minimal quasiconformal distortion under affine normalization. Optimal affine balancing is unique up to similarity and yields canonical harmonic mappings with centrally symmetric dilatation. Using a three-point support principle and a sharp
hyperbolic Jung
theorem, we establish universal sharp two-sided bounds for these invariants and provide their pseudohyperbolic reformulations. For canonically balanced mappings, we prove sharp second-order estimates for pre-Schwarzian derivatives, showing that affine normalization cancels leading-order conformal discrepancies. We develop an affine-stable hierarchical classification for families of harmonic quasiconformal mappings, including normality and an invariant ordering. This framework unifies extremal distortion problems and connects classical analytic function theory with the distortion theory of harmonic quasiconformal mappings.
Read the full abstract ▾
Computer Vision
TomoTransformer: Towards a Foundation Model for CT Reconstruction
AmirEhsan Khorashadizadeh and Benjamín Béjar
Supervised deep learning has advanced sparse-view tomographic reconstruction. However, conventional models, which typically map filtered back-projection (FBP) images or sinograms to clean reconstructions, are brittle under distribution shifts. Because they require retraining whenever projection counts and angles, detector resolutions, or data distributions change, their deployment in real-world applications remains limited. To address this, we introduce TomoTransformer, a
transformer-based architecture that treats each local filtered projection as an individual
token and predicts missing views via self-
attention. Crucially, TomoTransformer operates in a back-projection space that separates projections across spatial locations, making view interpolation geometrically well-posed and invariant to detector size. This design yields a single foundation model that can process any number of input projections, at arbitrary angular locations and detector dimensions, and query any number of target angles without retraining. Trained on a large-scale
dataset spanning diverse medical CT anatomies and natural images, TomoTransformer generalizes effectively across anatomies, materials, and resolutions. Extensive evaluations on several
benchmark sparse-view
datasets show that TomoTransformer significantly outperforms concurrent multi-purpose models like ViewTrans and matches or exceeds strong protocol-specific baselines, while remaining fully agnostic to the number of input and target projections. Furthermore, the model demonstrates robust zero-shot generalization on real experimental nanoscale
brain data collected from an X-ray synchrotron, showcasing its practical utility for real-world applications.
Read the full abstract ▾
Machine Learning
GraphVQ: Structure-Aware Autoregressive Decoding over Context-Quantized Graph Tokens
Yuxiang Yao and Zijun Zhao
Graph foundation models need a discrete
token representation, but casting a graph as a generatable
token sequence faces a structural obstacle: edges spanning beyond the serialization window cannot be emitted in one pass–so one-pass autoregressive generators systematically under-produce cycles–and a single global condition cannot tell candidate edges apart. GraphVQ removes both obstacles: node contexts–features plus a local edge mask under multi-order breadth-first serialization–are quantized into a shared codebook by a VQ-VAE with BCE-calibrated Bernoulli edge decoding, and a second-stage structure-aware decoder emits the global adjacency conditioned on
token-derived pair features, whose necessity over any global-summary condition is formalized in a scoped impossibility result. The tokenizer reconstructs node features at 0.86–0.99 accuracy and decodes local edges at AUROC >= 0.89 (ECE <= 0.007). Under one same-split protocol on four
datasets, pair conditioning improves orbit MMD 0.248 -> 0.174 on
PROTEINS and 3.4x on a ring stress test, and vanishes on a random-label control–the signature of attribute–
topology coupling–so the gain is claimed exactly where attributes carry edge-relevant signal. GraphVQ ranks first among learned generators on
PROTEINS, ties for first on SYN-COMM, and improves orbit MMD 2.7–17x over one-stage generation on three
datasets, with seed-level bootstrap intervals confirming the rankings are not seed noise; on MUTAG the unweighted edge target under-generates and is reported as such. These results locate the structural control of autoregressive graph generation in the granularity of the condition: pair-level
token context turns a quantized vocabulary into a usable capacity axis for distribution-faithful graph generation and future
token-level pretraining.
Read the full abstract ▾
Software Engineering
Independent Verification Paths Are Not Independent: A Case Study of Common-Mode Failure in a Satellite Catalogue Pipeline
Fabio Rovai
A common safeguard for a data pipeline is redundant computation: derive each published number by two routes built on different technology and refuse to exit when they disagree. We report one such gate failing, in a cross-catalogue integrity study of two open registers of Earth-orbiting objects. A gate comparing a set-based Python path with SPARQL queries over the emitted RDF graph printed ALL CROSS-CHECKS AGREE on seven counts. Three were wrong, one overstated more than fourfold (932 against 220). Both paths imported the same constants, which encoded a misreading of the source's status vocabulary, so the error was common-mode and the gate could not see it. We give the mechanism, an object-level ledger reconciling every figure, and three checks that go back to the source's documentation, measured on the defective code and on its correction. We then checked that correction against each object's phase history, held in a source file the pipeline never read. The correction was also wrong: 42 of its 261 disagreements are artefacts, and none of our three checks flagged them. Finally, in a controlled replication with three pinned models and tools disabled, 72 of 75 paths generated on request as independent checks computed the defective count, 29 of 30 even when the
prompt carried the source's own definitions of the codes. The evidence is one pipeline and one defect family. Within it, redundancy verified implementation, and the errors that reached publication were errors of meaning.
Read the full abstract ▾
Robotics
When to Adapt: Multi-Signal Domain Shift Detection for Efficient Training-Free Adaptation in Open-Vocabulary Segmentation
Michele Antonazzi, Alejandra C. Hernandez, José Araujo et al. (+2 more)
Robust and reliable perception is essential for autonomous
robots operating in real-world environments, particularly in long-term missions where environmental conditions may change significantly over time. Although recent advances in Visual Foundation Models (VFMs) have improved open-vocabulary semantic
segmentation, these models can still suffer from domain shift, which can significantly degrade performance if they are not adapted to the current environment. Training-free domain adaptation is a relevant paradigm for adaptation, consisting of adjusting the model online using lightweight adapters. Recent approaches apply this on a per-frame basis, which is impractical for deployments on resource-constrained robotic hardware. To tackle this, we propose a multi-signal domain shift detection method for training-free continual test-time adaptation (TF-CTTA) in open-vocabulary
segmentation. Our method leverages temporal coherence across consecutive frames by monitoring and combining complementary aspects of domain shift (visual change, adapter mismatch, and semantic drift) to trigger adaptation only when needed. We validate our approach on a
benchmark including indoor and outdoor environments and using real robotic data. We demonstrate that our approach maintains
segmentation accuracy while substantially reducing adaptations, making training-free adaptation practical and feasible for long-term, real-world robotic deployments.
Read the full abstract ▾
All authors Michele Antonazzi, Alejandra C. Hernandez, José Araujo, Olov Andersson, Patric Jensfelt
Audio and Speech Processing
SENSE: Semantic Neural Speech Synthesis from Brain Dynamics via Spatial Graph Encoding
Jisoo Park, Seonghak Lee, Hyojin Park et al. (+1 more)
Reconstructing
speech from non-invasive
brain signals offers a promising pathway for restoring communication in individuals who are cognitively intact but unable to speak. Existing
EEG-to-
speech approaches formulate this task as acoustic reconstruction, optimizing waveform fidelity while ignoring whether the generated
speech preserves high-level semantic content. In this work, we revisit this formulation and argue that
EEG signals carry not only acoustic but also semantic information. We identify two key limitations of prior methods: (1) the neglect of spatial relationships between
EEG electrodes, and (2) the failure to exploit the semantic structure of the N400 paradigm, where congruent and incongruent trials reflect distinct semantic processing. We propose SENSE(Semantic-
EEG Neural
Speech SynthEsis), which combines a graph-based
EEG encoder over electrode geometry with
EEG Semantic Conditioning (ESC), aligning
EEG to a pretrained semantic space using only congruent trials. On the N400
dataset, SENSE consistently outperforms prior methods on both acoustic and semantic metrics, and model-internal channel attribution suggests distributed reliance on auditory, sensorimotor, and centro-parietal regions, consistent with known
speech-perception neuroscience. In the unseen-subject setting, SENSE trained on only two subjects already surpasses the strongest baseline trained on all eighteen subjects in word error rate, and matches it on acoustic metrics with as few as eight subjects.
Read the full abstract ▾
All authors Jisoo Park, Seonghak Lee, Hyojin Park, Junseok Kwon
Systems and Control
Quantifying EnergyNet performance: a simulation-based framework for decentralized energy networks
Max Collins, Jonas Birgersson, Marc A. Weiss et al. (+10 more)
Power transmission and distribution infrastructure connects generation with consumers. Extreme weather, infrastructure failures, and geopolitical conflicts increasingly disrupt these connections, leaving large groups without electricity until service is restored. Local generation, storage, and controllable loads pose challenges but also enable new ways to organize energy systems. EnergyNet is a decentralized architecture of interconnected power electronic Energy Routers. Each coordinates local loads, generation, storage, and grid connections. Bidirectional Energy Links enable exchange between routers, while the Energy Protocol communicates needs and availability to coordinate transfers. EnergyNet can operate with or without the public grid. This paper develops a quantitative framework for modeling and evaluating EnergyNet. It represents
topology, prioritized loads, generation, storage, grid connections, links, and component limits. Receding-horizon optimization calculates energy allocation and exchange, supporting also disturbance and reliability analysis. A three-router case illustrates energy sharing and islanded operation. The framework then models a Stanford University campus microgrid and an off-grid microgrid in Half Moon Bay. Under low-
solar conditions, electric bus visits support the off-grid microgrid while preserving transport service. Finally, the paper models an EnergyNet under construction in Lund, Sweden, comprising ten buildings and 278 apartments. Full service is maintained in summer and winter and after the loss of one grid interface or Energy Link. Losing two links partitions the network; service in the island depends on local resources and repair time. Reliability screening identifies partitioning as rare but credible. Monte Carlo analysis shows full critical service in sampled summer conditions and median continuous critical service of approximately 92% in winter.
Read the full abstract ▾
All authors Max Collins, Jonas Birgersson, Marc A. Weiss, Jimmy Chen, Mark Z. Jacobson, Newsha K. Ajami, Daniel Kammen, Tomas Kåberger, Franklin Carrero-Martínez, Michael Menser, Henrik Ny, Lou Riordan, Jill G. Ferguson
Robotics
BlenDAgger: Blended Shared Control for Interactive Imitation Learning
Cailyn Smith, Geoffrey Sun, Henny Admoni et al. (+1 more)
Robot policies are frequently trained from human corrections, yet teleoperating a
robot to provide corrections is burdensome, and human demonstrators are not always optimal. We propose Blended DAgger (BlenDAgger), an approach for collecting data to train imitation learning policies by using shared control to blend the policy's and demonstrator's actions during interventions. By blending human and policy actions, we aim to improve the autonomous performance of manipulation policies. We validate our approach across five manipulation tasks, two in the real world and three in simulation. Our approach achieves higher autonomous performance by 30 or more percentage points on two real-world tasks compared to a typical human-gated correction approach (HG-DAgger). We also investigate the advantages of BlenDAgger that allow for higher autonomous performance, finding that BlenDAgger results in 57% smoother transitions between policy control and human interventions, and 14% higher trajectory similarity to the training data. In a user study (n=14) on two real-world tasks, we find that BlenDAgger results in faster data collection (BF=13.32), and we do not find a difference in subjective perceptions. These results show that blended shared control leads to higher autonomous performance compared to typical methods for fine-tuning
robot policies from fully teleoperated interventions.
Read the full abstract ▾
All authors Cailyn Smith, Geoffrey Sun, Henny Admoni, Zackory Erickson
Soft Condensed Matter
A Bioinspired Magnetothermally Triggered Capsule for Rapid Microrobot Release
Mengfan Zhang, Zike Chen, Zhihao Lv et al. (+3 more)
Capsules are attractive platforms for microrobot delivery, combining structural simplicity with protected delivery and rapid release remains challenging. Inspired by the latched spring mechanism of Impatiens balsamina seed dispersal, a magnetothermally triggered capsule integrates a phase-change hydrogel latch, a monostable cover that serves as an elastic spring, and a payload-carrying base. Mechanical analysis guides the cover design to avoid bistability after assembly and premature hydrogel latch rupture before triggering. During transport, the hydrogel constrains the inverted cover. Under an alternating
magnetic field, Fe3O4-mediated heating softens the hydrogel and induces cohesive failure, releasing stored energy to drive capsule opening and microrobot ejection. The capsule opens within 50 ms after hydrogel softening and ejects microrobots at an estimated initial velocity of 1.0 m/s. Demonstrations on a three-dimensional platform and ex vivo porcine stomach establish protected delivery, target-site opening, and post-release actuation.
Read the full abstract ▾
All authors Mengfan Zhang, Zike Chen, Zhihao Lv, Zheng Jia, Rui Xiao, Guoyong Mao
Earth and Planetary Astrophysics
TOI-6981 b: a transiting sub-Neptune at the water ice line
Madison G. Scott, Amaury H.M.J Triaud, Emily K. Pass et al. (+42 more)
We present TOI-6981 b, a cold (Teq= 175^{+6}_{-5} K, assuming an Earth-like albedo) sub-Neptune (Rb=2.51+/-0.11 R_oplus) orbiting on a 60.8 day orbit around around its M4-type host
star (R*=0.304 +/- 0.012 R_odot, M*=0.296 +/- 0.026 M_odot, Teff = 3215+/-58K), part of the SPECULOOS TEMPOS program. We validate the planetary nature of TOI-6981 b using ground-based follow-up photometry, statistical validation and high-resolution spectroscopy. At such cold temperatures, TOI-6981 b sits roughly at the location of the water ice line (approx 170 K). Given its orbital period and high transmission spectroscopy metric (TSM=56 +/- 10), it is one of the only
planets in the sparsely populated cold
planet regime amenable for atmospheric studies with JWST. Therefore, TOI-6981 b could. possibly become a crucial
benchmark in cold chemistry studies, tests of Hycean world hypotheses, and drawling parallels to the
Solar System's ice
giants. Additionally, given its current orbit, it may provide insight into planetary formation and migration at or beyond the water ice line.
Read the full abstract ▾
All authors Madison G. Scott, Amaury H.M.J Triaud, Emily K. Pass, Aleyna Adamson, Karen A. Collins, Khalid Barkaoui, Allyson Bieryla, Adam J. Burgasser, Mark E. Everett, Steve B. Howell, David W. Latham, Benjamin V. Rackham, Boris S. Safonov, Miquel Serra-Ricart, Abderahmane Soubkiou, Georgina Dransfield, Katharine Hesse, Pipa Fernandez de Diego, Emma Softich, Gregor Srdoc, Mathilde Timmermans, Jeroen Audenaert, Roi Alonso, Artem Y. Burdanov, Yasmin T. Davis, Fatemeh Davoudi, Julien de Wit, Shishir Dholakia, Elsa Ducrot, Michaël Gillon, Yilen Gomez Maqueo Chew, Matthew J. Hooton, Clàudia Janó-Muñoz, Emmanuel Jehin, Akanksha Khandelwal, Thomas R. Marriott, Peter P. Pedersen, Manuel Pichardo Marcano, Francisco J. Pozuelos, Rafael Rebolo-López, Manu Stalport, Alexander Venner, Selçuk Yalçinkaya, Francis Zong Lang, Sebastián Zúñiga-Fernández
Quantum Physics
Semiclassical spin-bath calculation of the nitrogen-isotope effect on NV$^-$ ensemble coherence in diamond
Chikara Shinei
The ratio $T_2$/$T_2^*$ of the Hahn-echo and Ramsey times of nitrogen-vacancy (NV$^-$) ensembles is independent of the nitrogen concentration [N] and is $\approx$ 16 in diamond of natural isotopic abundance ($^{14}$N), yet a recent $^{15}$N-doped ensemble magnetometer gives only 8.05 $\pm$ 0.18 in the single-
quantum convention. We ask whether the nitrogen
nuclear isotope alone can cause such a reduction. Using a semiclassical
spin-bath model with the Jahn-Teller-resolved P1 hyperfine tensor, we show that the fraction of P1 pairs that are hyperfine-degenerate, and hence free to flip-flop, is 1/4 for $^{14}$N (I = 1) but 5/16 for $^{15}$N (I = 1/2). Simulations over [N] = 0.1–100 ppm confirm that this shortens $T_2$ while leaving $T_2^*$ exactly unchanged: $T_2$($^{14}$N)/$T_2$($^{15}$N) = 1.118 $\pm$ 0.011, independent of [N] and 11$\sigma$ above unity. The measured contrast, 2.07 $\pm$ 0.25, is 3.8$\sigma$ larger, identifying resonant-channel counting as a real but partial contribution and setting a quantitative
benchmark for many-body calculations.
Read the full abstract ▾
Number Theory
Ax-Schanuel for the Drinfeld $j$-function
Gal Binyamini, Dmitry Novikov and Francesco Maria Saettone
We prove an analog of the Ax-Schanuel
theorem for the Drinfeld $j$-function in odd characteristic. Roughly speaking, if the graph of $\boldsymbol{j}\colon\Omega^n\rightarrow\mathbb{A}^n_{\mathbb{C}_\infty}$ and its derivatives has an atypical intersection $\mathcal{V}$ with an algebraic variety, then $\mathcal{V}$ projects to a weakly-special subvariety in $\mathbb{A}^n_{\mathbb{C}_\infty}$. More generally, we prove a positive characteristic analog of the differential Galois theoretic Ax-Schanuel
theorem of Blázquez-Sanz, Casale, Freitag and Nagloo. Our main
theorem for $\boldsymbol{j}$ follows by applying this result to a suitable foliation. A
theorem of Pink allows us to conclude that the special varieties in the sense of Blázquez-Sanz et al. in this context agree with the classical weakly-special subvarieties in the Drinfeld sense.
Read the full abstract ▾
Artificial Intelligence
XU-RS: Explaining Credal Width in Random-Set Language Models
David Achara, Maryam Sultana, Alexander D. Rast et al. (+1 more)
Uncertainty estimates tell us how unsure a model is, but not why. Without knowing which parts of an input influences a model's uncertainty, we cannot tell whether that uncertainty score depends on input features that are relevant for the task. We study this problem in randomset classifiers built using pretrained
language models. These classifiers assign probability to individual answers and to groups of answers, producing lower and upper probabilities for each answer; The difference between these probabilities, called credal width, is used to represent epistemic uncertainty about an answer arising from limited training data. We propose XU-RS, a framework that attributes an answer's credal width to the input
tokens (words or word pieces) supplied to a
language model. XU-RS uses Expected Gradients (a standard feature attribution method) to estimate how input
tokens contribute to credal width. The proposed framework is evaluated on a MedQA
dataset using SmolLM3-3B and Llama-2-7B models, demonstrating that setting the
embedding of a
token ranked highly by XU-RS to zero (zero-masking) causes larger changes in credal width than zero-masking randomly selected
tokens. In addition, we show that normalisation can cause other answer groups to influence an answer's width, reveal how
token attribution can mask numerical errors, and provide diagnostic checks to verify whether a
token ranked highly by XU-RS meaningfully explains model uncertainty.
Read the full abstract ▾
All authors David Achara, Maryam Sultana, Alexander D. Rast, Fabio Cuzzolin
Analysis of PDEs
Long time strong Sobolev instability for quasilinear NLS
Yuri Cacchiò, Filippo Giuliani, Felice Iandoli et al. (+1 more)
We prove long-time strong Sobolev instability for a class of quasilinear Schrödinger equations on $\mathbb T^2$ arising, via the Madelung transform, in the description of Euler-Korteweg capillary fluids. More precisely, for every $s\ge 7$, arbitrarily small $\mu>0$, and arbitrarily large $\mathcal K>0$, we construct a solution such that $$ \|u(0)\|_{H^s}<\mu, \qquad \|u(T)\|_{H^s}>\mathcal K $$ at some finite time $T$, for which we provide an explicit exponential upper bound. The instability mechanism is generated by a finite-dimensional resonant Toy Model associated with the cubic semilinear part of the equation, in the spirit of Colliander-Keel-Staffilani-Takaoka-Tao. The main technical difficulty is to justify this dynamics over the long time scale of the energy transfer in the presence of quasilinear derivative losses. We achieve this by combining a weak Birkhoff normal form with paradifferential calculus, microlocal symmetrization, and modified energy estimates. Our result provides a rigorous quasilinear realization of a resonant energy-transfer mechanism, and suggests a connection with weakly turbulent dynamics in capillary fluids.
Read the full abstract ▾
All authors Yuri Cacchiò, Filippo Giuliani, Felice Iandoli, Raffaele Scandone
Statistical Mechanics
Quantum Statistical Thermal Engine at the BCS-BEC crossover
Santiago Henríquez Lira, Felipe Isaule, Martín HvE Groves et al. (+3 more)
We propose a
quantum heat engine based on a two-component
Fermi gas with s-wave contact interaction, operating across the BCS-BEC crossover. The work is extracted from the statistical properties of the gas, which are controlled by the interaction, rather than relying solely on compression and expansion stages. Using the functional renormalization group formalism, we obtain the equation of state in a non-perturbative form along the crossover, encompassing the superfluid-normal phase transition. The cycle combines isentropic density strokes, isochoric thermalization, and isothermal interaction sweeps. This construction makes it possible to integrate features of both Otto and Carnot cycles, in which the system simultaneously saturates both efficiency limits without the net work vanishing. In the absence of density variations, the cycle reduces to a statistical Stirling-like engine, in which the work generated arises exclusively from the interaction, achieving efficiencies up to $38\%$. A pronounced asymmetry emerges through the crossover, giving rise to distinct operating regimes depending on the trajectory followed in the phase diagram. Consequently, the same architecture can be tuned to function as an engine, refrigerator, accelerator, or heater. These findings highlight pairing correlations as a versatile thermodynamic resource for
quantum heat machines.
Read the full abstract ▾
All authors Santiago Henríquez Lira, Felipe Isaule, Martín HvE Groves, Francisco J. Peña, Patricio Vargas, Thomás Fogarty
Robotics
Credit-Guided Policy Improvement for Test-time Adaptive Vision-Language Navigation
Yang Li, Sijia Zhang, Yihan Li et al. (+4 more)
Test-time adaptation for
vision-
language navigation (TTA-VLN) enables pretrained policies to adapt online to unseen environments using only test-time observations and interaction history. However, distribution shifts can distort local action preferences and lead to off-course decisions. Existing methods rely on predictive uncertainty, trajectory-level feedback, or accumulated adaptation experience to correct such deviations. These signals, however, do not directly reveal whether an executed action supports instruction-guided progress toward the goal. Moreover, a plausible corrective signal does not guarantee a reliable policy update. The key challenge is thus twofold: identifying interactions that support goal-directed improvement and determining whether the resulting updates are worth retaining. We observe that each executed action induces an immediate observation transition, providing evidence of its local consequences. Based on this insight, we propose Credit-Guided Policy Improvement (CGPI), which recovers signed, reference-relative decision credit from action-induced observation transitions without external outcome feedback. With the pretrained
navigation policy frozen, CGPI uses this credit to propose lightweight adaptation updates and verifies them against prior credit-supported interactions. Updates are retained only when supported and rolled back otherwise. CGPI achieves consistent gains across the evaluated VLN
benchmarks and
navigation backbones, while qualitative
robot trials further illustrate the feasibility of zero-shot sim-to-real transfer.
Read the full abstract ▾
All authors Yang Li, Sijia Zhang, Yihan Li, Aming WU, Zihao Zhang, Ziju Han, Yahong Han
Artificial Intelligence
FOCUS: Training-Free Decision-Preserving Context Compression for LLM Agents
Shantanu Dixit, Anson Bastos, Xuchao Zhang et al. (+2 more)
LLM
agents accumulate interaction histories that grow linearly with task length, causing quadratic inference cost scaling and performance degradation from
attention dilution. Existing context-compression methods learn what to discard offline: by contrastively optimizing guidelines, distilling compressors, or training compression policies. This incurs a substantial cost. Further, the compression policy is learned a priori and is not dynamically conditioned on the evolving test-time trajectories. In this paper we ask a complementary question: Which past interactions causally shape the
agent's future decisions? We recast context compression as a causal decision preservation problem over discrete interaction units and introduce FOCUS, a training-free context compression framework that operates entirely at test time. Our method requires no offline data collection or fine-tuning, and is architecture-agnostic, attaching to any closed-API frontier model as a modular compression layer. We evaluate FOCUS on diverse agentic
benchmarks including API and tool-calling, QA, web domain and multi-turn dialogue. Our method establishes new state of the art performance, cutting peak context by up to 48% and dependency by 73% while improving task success by up to 8.9 percentage points over uncompressed execution.
Read the full abstract ▾
All authors Shantanu Dixit, Anson Bastos, Xuchao Zhang, Chetan Bansal, Saravan Rajmohan
Optimization and Control
Common Optimal Supports and Bottleneck Hierarchies in Truncated Cones
Walid Ben-Ameur and Alessandro Maddaloni
{Let \(K\subseteq\mathbb R_+^E\) be a closed convex cone and \(P=K\cap[0,1]^E\). For a linear objective with positive optimum, we study the intersection of the coordinate supports of all optimal solutions.} We prove that this common support is a transversal of a natural clutter associated with the positive directions of the cone and introduce a decreasing hierarchy of bottleneck families that stabilizes after at most \(|E|-1\) levels. In the polyhedral setting, the first level consists exactly of the sets carrying feasible upper-bound dual multipliers, while every set carrying an optimal multiplier belongs to the stabilized family. {The breadth of the join-semilattice generated under union by the zero sets of the nonzero vertices gives an objective-independent upper bound on the universal stabilization depth;} this bound is exact for simplicial truncations. Every possible depth occurs, and the \(|E|-1\) bound is sharp. Membership in the second level is co-NP-complete and fixed-parameter tractable in the size of the candidate bottleneck. Under upper-box-integrality and the integer decomposition property, the hierarchy collapses at level two, and its stabilized members are exactly the carriers of optimal upper-bound multipliers. When \(P\) is integral, the common optimal support is the union of the supports of all optimal dual solutions. Applications include bipartite matchings, maximum-order cycle subdigraphs, binary circulations, balanced hypergraph matchings, interval hypergraphs, and maximum-weight closures.
Read the full abstract ▾
Artificial Intelligence
Rational Clarification by Assistive Agents via Value-of-Information Reasoning
T. Duy Nguyen-Hien, Yee Whye Teh, Wee Sun Lee et al. (+1 more)
Users of
language-based assistive
agents often make ambiguous requests. In response, an assistant can either directly act on its interpretation of the request — risking misalignment with the user — or ask a clarifying question. Which option is the most safe and helpful? A common approach is to ask questions that minimize uncertainty about the user's intent until a threshold is reached. However, this neglects the impact of uncertainty reduction on downstream performance, the costs of asking versus acting immediately, and the possibility that users may provide corrections without being asked. To navigate these trade-offs, we introduce Rational Enquiry via Value-of-Information
Reasoning (REVOIR). REVOIR makes clarification decisions via inference-time
reasoning about the value-of-information of a question, which captures the expected improvement in task reward due to the answer received. In two assistive tasks — ambiguous question answering (CondAmbigQA) and preference-aligned household task planning (ADAPT) — we show that REVOIR achieves greater success with fewer questions than approaches based on prompting, chain-of-thought, fine-tuning, or information gain, improving preference satisfaction on ADAPT by 13-15% over a fine-tuned clarification policy while requiring no training and asking five times fewer questions. Furthermore, when the assistant can receive cheap user corrections after acting, REVOIR naturally infers that asking questions is not always efficient, demonstrating the adaptivity of our approach. In contrast, we find that vanilla
reasoning agents fail to adaptively clarify user requests, and request fewer clarifications as
reasoning effort increases.
Read the full abstract ▾
All authors T. Duy Nguyen-Hien, Yee Whye Teh, Wee Sun Lee, Tan Zhi-Xuan
Machine Learning
ReLMem: Learning Recurrent Memory for Longitudinal EHR Modeling
Zijie Meng, Xiwei Dai, Yingying Zhang et al. (+3 more)
Longitudinal electronic health record (EHR) modeling requires integrating new visits with an expanding patient history. Yet the continual accumulation of clinical information imposes increasing computational and
memory costs on large
language models (LLMs) when they process and retain complete patient histories. A practical alternative is visit-wise recurrent compression, which incorporates each incoming visit into a compact, continually updated patient
memory. However, under a fixed
memory budget, successive updates must integrate new information without progressively losing critical historical evidence needed to subsequent tasks. To address this challenge, we introduce Recurrent Longitudinal
Memory (ReLMem), a framework that learns to maintain fixed-capacity patient
memory for efficient downstream prediction with a frozen LLM. ReLMem equips this LLM with lightweight compression adapters to recurrently update the
memory from its previous state and each incoming visit, without rereading earlier records. Specifically, we develop a multi-granularity optimization strategy to preserve task-relevant information throughout recurrent updates and support downstream prediction from the final
memory. The intermediate supervision aligns
attention outputs from compressed
memory and the full history under identical queries, while prediction supervision minimizes cross-
entropy with ground truth answers conditioned on the final
memory. On EHR-based medication prediction, ReLMem approaches the F1 scores of full-history baseline while reducing average retained historical storage by 97.1%. Under the same
memory budget, it improves macro- and micro-F1 over the strongest compressed-
memory baseline by 4.66 and 4.75 percentage points, respectively. These results highlight the value of learning recurrent patient
memory for efficient longitudinal EHR modeling.
Read the full abstract ▾
All authors Zijie Meng, Xiwei Dai, Yingying Zhang, Jian Wu, Xian Wu, Zuozhu Liu
Sound
Learning as Deepfakes Evolve: RF-Prompt for Continual Audio Deepfake Detection
Yuankun Xie, Xiaoxuan Guo, Xiaopeng Wang et al. (+3 more)
Continual audio deepfake detection requires learning newly emerging deepfake methods while retaining discrimination of previously encountered
speech. Existing
dataset-incremental evaluation changes both real-
speech domains and deepfake mechanisms, making their effects difficult to distinguish. We construct five task organizations over identical training, development, and evaluation pools to study these factors under a controlled sample budget. Our proposed Real-Anchored Mechanism-Incremental (RAMI) protocol reflects the practical setting in which available real
speech provides a recurring mixed-domain reference while new deepfake mechanisms arrive incrementally. We further propose RF-
Prompt, an asymmetric continual
prompt-learning method that preserves reusable real-
speech knowledge through a shared real
prompt and expands mechanism-specific knowledge through inherited fake experts with orthogonal residuals. Input-adaptive soft fusion combines the accumulated experts into a fixed number of injected
tokens without requiring task identity at inference. On RAMI, RF-
Prompt achieves 10.110% average EER and 10.370% pooled EER, outperforming all evaluated continual-learning baselines. Across the five controlled protocols, RAMI yields the lowest common-average and pooled EER. Component ablations, limited-data experiments, and cross-backbone evaluations further validate the proposed design.
Read the full abstract ▾
All authors Yuankun Xie, Xiaoxuan Guo, Xiaopeng Wang, Siqing Qin, Shaole Li, Kong Aik Lee
General Relativity and Quantum Cosmology
Deformation Quantization of Schutz Quantum Cosmology: Relational Time, Constraints, and Physical States
Raphael Fracalossi
A controlled correspondence is established between reduced relational dynamics and extended constrained quantization in an FLRW cosmology coupled to a Schutz perfect fluid. A canonical transformation in the matter sector yields a deparametrized constraint of the form $\mathcal C=p_T+H$, with $T$ serving as an internal time. In the extended Weyl–Wigner formulation, the
quantum constraint is imposed bilaterally. Its antisymmetric part gives the exact Moyal
evolution generated by the reduced
Hamiltonian, while its symmetric part imposes an independent
quantum shell condition. For a fixed self-adjoint realization of $\hat H$, every reduced density operator supported on its negative spectral subspace admits an explicit distributional extension. A spectral dyad with energies $E$ and $E'$ is mapped to a clock Wigner distribution supported at $p_T=-(E+E')/2$, with relational phase determined by $E-E'$. The map preserves off-diagonal coherences, satisfies both
star constraints, and has the reduced Wigner state as its $p_T$ marginal; it is therefore injective on the admissible positive-density sector. A flat
dust model with $q=1/2$ and Dirichlet data provides an explicit example: its reduced spectrum is $(-\infty,0]$, so the admissible sector contains all normalizable reduced states. The correspondence applies to a fixed constraint representative and
quantum realization.
Read the full abstract ▾
Analysis of PDEs
Formation of shear bands through discrete localized plastic increments
Heiner Olbermann and Hidde Schönberger
We study rate-independent systems in which the
evolution is obtained by incremental minimization of elastic energy and one-homogeneous dissipation under constraints on the admissible inelastic increments. Such constraints arise naturally in several models of inelastic materials, including shear-transformation-zone models, intermittent plasticity, and quantized plasticity, where
evolution proceeds through prescribed elementary activation events rather than arbitrary plastic increments. We establish the existence of a continuous-time limit for the corresponding constrained incremental scheme under suitable assumptions. As an application, we formulate a simplified shear-transformation-zone model consisting of localized activation events on a
lattice embedded in a two-dimensional elastic medium. We prove that this model generates shear-band formation through the successive activation of neighboring transformation zones. More precisely, we show that every admissible
evolution remains localized on a growing line segment in the infinite-domain setting and that the same mechanism persists on sufficiently large bounded domains. These results provide a rigorous mathematical framework for constrained activation models and show that restrictions on admissible elementary events alone constitute a mechanism for localization in rate-independent systems.
Read the full abstract ▾
Robotics
RoboHarn-Evo: Evolving Hierarchical Physical Knowledge for Self-Improving Robotic Manipulation
Shifeng Bao, Fanding Huang, Yihan Lin et al. (+7 more)
Vision-
language models can coordinate long-horizon
robot manipulation, yet successful task
reasoning still depends on whether local physical interactions produce the intended effects. We study how repeated interaction can improve this capability without updating the base model. We introduce RoboHarn-Evo, a dual-loop harness that evolves Hierarchical Physical Knowledge (HPK) from physical experience. HPK couples two levels of reusable knowledge: Task Knowledge captures which subtask should be executed and when it is complete, while Action Knowledge captures object-relative
geometric strategies and their physical effects. During execution, the
agent retrieves knowledge at the corresponding decision level and grounds it in the current scene under the task goal. Across episodes, physical feedback is used to revise historical knowledge, update its applicability, and organize reusable entries for subsequent
retrieval. Experiments on RMBench show that HPK improves average success by up to 24.2 percentage points across different
agent models. With 80 interaction rollouts, held-out success rises from 48.3% to 75.0% for GPT-5.5 and from 70.0% to 88.3% for GPT-6. RoboHarn-Evo also resolves over 83% of historical knowledge errors while retaining 95.8% of valid knowledge, and transfers zero-shot from RMBench to RoboDojo with gains of 35.0 and 25.0 percentage points. These results demonstrate that physical interaction can be accumulated into reusable knowledge for improving subsequent manipulation.
Read the full abstract ▾
All authors Shifeng Bao, Fanding Huang, Yihan Lin, Youhe Feng, Guanlin Li, Chen Zhao, Yang Li, Jiawei He, Cheng Chi, Jing Zhang
Computer Vision
FedSocket: Recipient-Executable Knowledge Exchange for Heterogeneous Multimodal Federated Learning
Xinyuan Zhao
Federated knowledge must remain usable by recipients with different modalities, private architectures, and tasks. We present FedSocket, which makes recipient execution a design requirement of the exchanged model. A shared Q combines recipient-computable inputs, task-owned outputs, and ownership-aware aggregation, connecting heterogeneous private models through a common prediction interface. Private models teach local Q copies; the returned Q supports local learning and Joint inference, with only Q parameters and counts exchanged. Across six
datasets, FedSocket improves missing-modality recipient accuracy over Local by 14.44 and 15.51 percentage points on MELD and UCF-51. Under matched inference capacity, Joint exceeds independent ensembles by 11.06 points in UCF-51 accuracy and 4.87 points in mean bidirectional Flickr30k R@1. Joint also improves over Q alone on all four heterogeneous endpoints, demonstrating the value of combining local and exchanged predictions. Teacher controls, sharing-path interventions, and component factorials identify the roles of supervision, sharing, and deployment. FedSocket makes exchanged knowledge directly usable from
federated training to recipient inference.
Read the full abstract ▾
Computer Vision
TReVS: Integrating Textual Relevance and Visual Saliency for Efficient Vision-Language Model Token Pruning
Jing Wang, Zhiping Wu, Dongdong Ren et al. (+3 more)
Vision-
Language Models (VLMs) excel at visual understanding and
reasoning but often incur substantial inference costs due to the large number of visual
tokens. Recent visual
token pruning methods increasingly follow a two-stage paradigm: they first remove visually redundant
tokens after the
vision encoder and then discard
tokens irrelevant to the textual query within the Large
Language Model (LLM). However, since the first stage typically relies solely on
vision-encoder saliency, it may prematurely eliminate query-relevant
tokens, depriving the subsequent text-guided stage of critical visual evidence. Our empirical analysis shows that incorporating query guidance into first-stage
pruning better preserves task-relevant evidence and consistently improves performance over
vision-only saliency-based
pruning. We further find that high-variance
attention heads are more sensitive to the textual query and yield more discriminative text-to-
vision attention signals for second-stage
pruning. Motivated by these findings, we propose TReVS, a training-free framework that combines textual relevance with
vision-encoder saliency for pre-LLM
pruning and leverages high-variance
attention heads to remove task-irrelevant
tokens at shallow-to-intermediate layers of the LLM. On LLaVA-1.5-7B, TReVS retains 92.8% of the unpruned baseline performance while
pruning 94.4% of visual
tokens, outperforming prior state-of-the-art methods.
Read the full abstract ▾
All authors Jing Wang, Zhiping Wu, Dongdong Ren, Youfang Han, Wei Zhao, Wenbin Li
Metric Geometry
Tensor valuations on Sobolev spaces
Tian Gao and Dan Ma
A complete classification is established for continuous, SL($n$) contravariant, and translation invariant tensor valuations defined on the Sobolev space $W^{1,p}(\mathbb R^n)$. When these valuations are further assumed to be homogeneous, the classification reveals that they are precisely the Fisher information tensors, which constitute a higher-order generalization of the Fisher information
matrix.
Read the full abstract ▾
Machine Learning (Statistics)
Ornstein-Uhlenbeck Is Hard to Beat, Yet Superlinear Drift Ships Lower Transport Costs
Attila Lovas and Lóránt Nagy
Brešar and Mijatović \cite{bresar2025} show that Ornstein–Uhlenbeck
diffusion is hard to beat in forward convergence under assumptions that exclude superlinear drift. We instead test superlinear Langevin
diffusions for score-based image generation, computing their conditional scores numerically from a Fokker–Planck equation. In our experiments, the superlinear models beat the Ornstein–Uhlenbeck baseline on empirical Wasserstein distance across nearly the entire tested grid and show less variation across
diffusion horizons. The “hard to beat” verdict of \cite{bresar2025} thus fails to be universal.
Read the full abstract ▾