Artificial Intelligence
Compositional Reasoning in Language Models under Reinforcement Learning Post-Training
Yu He, Yingxi Li, Yifei Wang et al. (+1 more)
Compositional
reasoning is critical for real-world problem solving: since training data is necessarily limited, models must generalize by composing learned skills in new ways. While post-training methods such as reinforcement learning (RL) have substantially improved the
reasoning abilities of
language models (LMs), their effects on compositional
reasoning remain less well understood. We propose a dependency-graph framework to formalize compositional
reasoning, yielding three levels of compositionality with increasing complexity. Empirically, we instantiate this framework with data-structure tasks, which provide deterministic reward computation and clear compositional structure. We find a consistent decomposed-to-composed asymmetry: decomposed-skill training does not reliably transfer to composed tasks, whereas composed-task training transfers more readily back to decomposed tasks. We provide theoretical explanation for this asymmetry, and further evaluate compositional generalization under length extrapolation, structural distribution shift, and transfer to tasks requiring unseen skills. Finally, we present a pilot study on real-world tool-calling
benchmarks, showing preliminary evidence that the decomposed-to-composed asymmetry can extend to practical settings.
Read the full abstract ▾
All authors Yu He, Yingxi Li, Yifei Wang, Ellen Vitercik
Differential Geometry
Complete Self-Shrinkers in Arbitrary Codimension with Positive Constant Scalar Curvature and $S\le 1$
Shunzi Guo
Let \(X:M^n\to\mathbb R^{n+p}\) be an \(n\)-dimensional complete self-shrinker in arbitrary codimension \(p\ge 1\). Suppose that the scalar
curvature \(R\) is a positive constant and that the squared norm \(S\) of the second fundamental form satisfies \(S\le 1\). We prove that \(S\equiv 1\) and that \(X\) is isometric to either the round sphere \(S^n(\sqrt n)\) or the standard generalized cylinder \(S^k(\sqrt k)\times\mathbb R^{n-k}\), \(2\le k\le n-1\). In particular, no new higher-codimension examples occur under these assumptions. The proof uses an algebraic lemma which yields a uniform positive lower bound for the Bakry–Émery
Ricci curvature, together with the comparison
theorem of Wei–Wylie, the equivalence of finite Gaussian volume and
polynomial volume growth due to Cheng–Zhou, and the gap
theorem of Cao–Li.
Read the full abstract ▾
Computer Vision
ParticleSplat: Self-supervised Object-centric Latent Particle Splatting
Lyuxing He, Daniel Guo, Elizabeth Terveen et al. (+3 more)
We present ParticleSplat, a self-supervised object-centric learning method that decomposes scenes into a set of latent ”particles” representing semantic entities through feedforward 3D Gaussian Splatting. Building on the Deep Latent Particles (DLP) framework, which represents images as a set of particles with attributes such as position, scale, and visual appearance, we address a key limitation of DLP: its inherently 2D nature, which prevents explicit 3D spatial and
geometric reasoning that are critical for downstream tasks such as robotic manipulation. Leveraging the structural similarity between latent particles and 3D Gaussian primitives, we introduce a 3D latent particle space trained with a novel view synthesis objective. Our model jointly encodes multiple views with camera poses into a shared 3D object-centric latent space, then transforms particles into particle-aligned 3D Gaussians whose composition reconstructs the full scene. On simulated and real-world
datasets, we show that this formulation inherently learns object masks without supervision and supports controllable 3D scene editing, such as moving objects by modifying particles in the latent space. We further establish that the learned 3D representation improves downstream performance on robotic manipulation tasks.
Read the full abstract ▾
All authors Lyuxing He, Daniel Guo, Elizabeth Terveen, Deepak Pathak, David Held, Tal Daniel
General Relativity and Quantum Cosmology
James Buda, Andrew Laeuger and Yanbei Chen
Gravitational-wave (GW)
memory is a permanent change in spacetime geometry induced by a burst of gravitational waves. It is theoretically interesting for its connection to asymptotic
symmetries, but has not yet been observed. The upcoming
Laser Interferometer Space
Antenna (LISA) is predicted to detect
memory directly in loud events from supermassive black hole (SMBH) mergers. We assess the prospects for the cumulative effect of
memory contributions as a
stochastic GW
memory background (SGWMB). Using two SMBH population models, we recover previously obtained single-event SNR forecasts and compute the power spectrum of the unresolved background, using numerical-relativity waveforms for modeling the
memory signal during each event. We find the predicted spectrum is significantly smaller than ones found by modeling the
memory waveform with a Heaviside function, with expected SNRs of order $O(0.1-10)$. We further find that our models favor a highly non-Gaussian ("popcorn") spectrum, demonstrating that the background may not appear as the union of continuous, overlapping signals, but rather as a collection of intermittent bursts. In both measures, we demonstrate that any current prediction of the spectrum is heavily dependent on the choice of population model. We interpret these findings in the light of the future LISA mission and emphasize their importance for properly handling
memory within the LISA global fit.
Read the full abstract ▾
Solar and Stellar Astrophysics
Energy redistribution during 2D oscillatory reconnection
Luiz A. C. A. Schiavo, Jordan Talbot, Gert J. J. Botha et al. (+4 more)
Magnetic reconnection is a fundamental energy release mechanism in
solar and space
plasmas. However, the distribution of released
magnetic energy among kinetic and internal (thermal) energy during time-dependent reconnection remains largely unexplored. This work aims to study energy conversion in magnetohydrodynamics via oscillatory reconnection (OR) at a 2D
magnetic null point. Using a shock-capturing code, we identify three families of shocks: termination shocks, slow-mode deflection shocks, and Petschek-like shocks (even with uniform resistivity). We measure the time-dependent reconnection rate, finding a peak of $R \approx 0.1$, consistent with a fast reconnection regime. For the first time, we track the energy conversion pathways between
magnetic, kinetic and internal energy across the time-dependent reconnecting system. We find energy exchange is strongly region dependent, with shock locations dominating over the
diffusion region. In the classical incompressible steady-state reconnection models, the
magnetic energy is fully converted into internal and kinetic energy. In contrast, we find that in compressible time-dependent reconnection, the interchange between kinetic and internal energy dominates the system (due to the inclusion of
plasma compressibility).
Read the full abstract ▾
All authors Luiz A. C. A. Schiavo, Jordan Talbot, Gert J. J. Botha, Ben Snow, Daniel Johnson, Madeleine L. Combrinck, James A. McLaughlin
Robotics
Pose-aware Legged Robot Semantic Exploration with Omnidirectional Perception in Confined Unknown Environments
Xiaoyang Zhan, Shiyu Chen and Kenji Shimada
Semantic exploration in confined environments requires both environment mapping and detailed observation of target objects. For ground
robots, limited sensor vertical fields of view and restricted standoff distances can leave upper object surfaces unobserved from planar viewpoints. Body tilting can improve coverage, but additional observations and posture transitions increase mission time. To address this trade-off, we present POSE, a pose-aware semantic exploration system that exploits a legged
robot's intrinsic body pitch and roll with omnidirectional camera-
LiDAR perception. The proposed pose-aware viewpoint sampling module selects body postures from partial object maps according to expected coverage gain, while aim-aligned execution reduces unnecessary body reorientation. Further, we introduce an object-centric viewpoint
pruning strategy assisted by a
vision-
language model (VLM), which uses persistent observation history and bird's-eye-view (BEV) maps to reduce redundant inspection visits. The resulting semantic viewpoints are combined with
geometric exploration viewpoints in a global exploration planner. Simulations show that POSE improves final target-surface coverage by 8-10 percentage points over the planar planning baseline while reducing exploration time by 17-32%, and achieves the highest mean object coverage AUC among the evaluated baselines. Real-world experiments with a legged
robot carrying an omnidirectional camera-
LiDAR suite in a machine shop further demonstrate the system's applicability. These results support adaptive body-posture planning for improving the coverage-efficiency trade-off in legged
robot semantic exploration. We plan to release the code for community benefit in the future.
Read the full abstract ▾
Astrophysics of Galaxies
Extreme Enrichment at Cosmic Dawn: Three FeII-Strong Quasars at z > 6 from the JWST Aether survey
Anniek J. Gloudemans, Emanuele Paolo Farina, Roberto Decarli et al. (+22 more)
We present three outliers on the
quasar main sequence at z>6, as revealed by JWST/NIRSpec IFU observations from the Aether survey, a program targeting ~200
quasars at
cosmic dawn. The
quasars, J0216-5226, J0320-3521, and J1429+5447, exhibit exceptionally strong FeII emission and weak H$\beta$ and [OIII] lines, indicating rapid chemical enrichment within the first Gyr of
cosmic history. Their FeII strengths ($R_{\rm{FeII},\lambda4570} \equiv \rm{EW}_{\rm{FeII}}/\rm{EW}_{\rm{H}\beta}$) range between ~2.5-3.6, placing them among the top ~3% of SDSS
quasars at ($0\lesssim z\lesssim 1$). The H$\alpha$ emission lines of J0216-5226 and J1429+5447 imply black hole masses of $\sim1-7\times10^9$ M$_{\odot}$ with $\lambda_{\rm{Edd}}\sim0.1-0.6$, comparable to typical high-z
quasars, although virial black hole masses may be systematically overestimated in strong FeII emitters. In contrast, J0320-3521 hosts a lower mass black hole of $\sim0.2-2\times10^9$ M$_{\odot}$, suggesting possible super-Eddington rate
accretion with $\lambda_{\rm{Edd}}$~0.5-6.0. The IFU data of J0320-3521 reveal a distinct emission-line region at a projected distance of ~1.5 kpc, consistent with either an interacting companion or ionized gas associated to the
quasar. We further identify a complex environment around J1429+5447, including two candidate companion
galaxies or extended ionized gas structures. These
quasars with bright emission-line regions, J0320-3521 and J1429+5447, are
radio-loud sources, while J0216-5226 remains undetected in
radio ($R_{5\rm{GHz,} \,3\sigma}\lesssim4$). Together, these results confirm previous findings that highly enriched, rapidly growing black holes were already in place at
cosmic dawn. Despite the small sample size, our results further highlight the diversity in environment and
radio-jet properties among
quasars that otherwise share the same extreme region of the
quasar fundamental plane.
Read the full abstract ▾
All authors Anniek J. Gloudemans, Emanuele Paolo Farina, Roberto Decarli, Klaudia Protušová, Fabrizio Arrigoni Battaia, Eduardo Bañados, Aaron J. Barth, Silvia Belladitta, Manuela Bischetti, Hyunseop Choi, Dominika Ďurovčíková, Anna-Christina Eilers, Simona Gallerani, Thales A. Gutcke, Luca Ighina, Brian C. Lemaux, Federica Loiacono, Yoshiki Matsuoka, Chiara Mazzucchelli, Silvia Onorato, Hyewon Suh, Fabian Walter, Feige Wang, Julien Wolf, Jinyi Yang
Signal Processing
Closed-form weak-wave propagation pressure amplitudes for arbitrary power-law attenuation
Esteban Avilés, Michael Oelze, Roberto Lavarello et al. (+1 more)
Biological tissues exhibit power-law frequency dependence of the attenuation coefficient. We derive closed-form weak-wave expressions for fundamental, second-, and third-harmonic pressure amplitudes of monochromatic acoustic plane waves in homogeneous media with arbitrary power-law attenuation to extend classical expressions for quadratic attenuation. For linear frequency dependence of attenuation, one-dimensional simulations using the k-Wave k-space pseudospectral solver yielded 35-42% (classical) and 0-2% (proposed) normalized root-mean-square errors for fundamental-band depletion. This extension supports improved fundamental-band B/A estimation for quantitative ultrasound tissue characterization. A proof-of-concept through-transmission application in phantoms with nearly linear attenuation yielded experimental B/A estimation errors of 23.7-52.7% (classical) and 4.0-5.9% (proposed).
Read the full abstract ▾
All authors Esteban Avilés, Michael Oelze, Roberto Lavarello, Andres Coila
Information Theory
Source Entropy-Guided Adaptive Transmission for Communication-Driven Multi-View Sensing
Mingjie Yang, Guangming Liang, Dongzhu Liu et al. (+3 more)
Communication-driven multi-view sensing relies on routine communication transmissions for sensing acquisition, while the resulting sensing data at distributed devices must be uploaded to an edge server under limited communication resources. This creates a unique coupling between sensing acquisition and edge inference: the communication interval determines the source information, whereas the uplink condition determines how much information can be delivered to the server for sensing inference. To account for this coupling, we propose a source
entropy-guided adaptive transmission framework. Specifically, we characterize the
entropy of packet-triggered channel state information (CSI) as a function of the communication interval using a multi-output Gaussian process. The resulting analytical bound is compared with the available bit budget, determined by the transmission rate and latency requirement, to select between original-data and task-oriented transmission. For task-oriented transmission, we formulate the communication-constrained inference problem based on the information bottleneck and decompose it into adaptive distributed encoding and multi-view inference (ADE-MI), which avoids alternating optimization between the devices and the edge server. Experiments on the Widar3.0 multi-view CSI gesture recognition
dataset show that the analytical bound closely follows the normalizing-flow numerical estimate, while ADE-MI outperforms task-oriented
benchmarks under the same bit budget and the proposed framework further improves recognition accuracy under time-varying channels.
Read the full abstract ▾
All authors Mingjie Yang, Guangming Liang, Dongzhu Liu, Lei Zhang, Xiaonan Liu, Kaibin Huang
Cryptography and Security
Beyond Private Training: The New Landscape of AI Privacy
Sean Culatana and Kang Li
Retrieval-augmented systems increasingly rely on vector indexes that may retain deleted items in their search graph. Existing deletion interfaces can prevent deleted identifiers from appearing in returned results while still computing distances to their
embeddings during graph traversal. We formalize this distinction as output safety versus traversal safety, and introduce TSD-AUDIT, a framework for
auditing and enforcing traversal-safe deletion in graph-based approximate nearest-neighbor
retrieval. On Faiss IndexHNSWFlat, native filtering leaves the number of distance computations unchanged relative to unfiltered search; at a 70% deletion rate, trace-faithful replay detects deleted-vector scoring in all 100 audited queries. Code inspection of hnswlib's mark_deleted path reveals the same scoring-before-liveness pattern. TSD-AUDIT enforces an alive-before-scoring invariant, repairs connectivity using only live candidates, and emits per-query scored-trace certificates that an independent verifier can check against the deletion snapshot. Under region-targeted deletion, TSD-AUDIT improves Recall@10 over native filtering by 4.3–42.2 percentage points across deletion fractions from 0.5 to 0.9, while remaining comparable under random deletion. These results show that output-only deletion audits can miss process-level exposure:
auditing deletion in vector
retrieval requires accounting for the vectors scored during search, not only the identifiers returned.
Read the full abstract ▾
General Economics
The Behavior of Inventories over the Business Cycle: Evidence across Levels of Development
Juan Manuel Rodriguez Repeti and Danilo Rogelio Trupkin
This paper documents how inventory dynamics vary across levels of development and how they respond to real and financial shocks. Using a balanced panel of 72 small open economies over 1993-2022, spanning advanced to low-income economies, we show that inventories are strongly procyclical in advanced economies but become acyclical at lower income levels, while
volatility and persistence rise as development declines. Using local projections, we estimate inventory responses to a long-run productivity shock and to a financial shock identified from the spread between U.S. corporate bond yields and Treasury yields. Productivity shocks generate temporary inventory accumulation across all groups, with smoother adjustment in advanced economies and more irregular cycles in emerging and developing economies. Financial shocks trigger decumulation followed by a rebound in higher-income groups, but a delayed, unreversed decline in low-income economies.
Read the full abstract ▾
High Energy Physics (Theory)
Notes on de Sitter space
Beatrix Mühlmann
We discuss classical and
quantum features of spacetimes with a positive cosmological constant. After introducing basic aspects of de Sitter cosmology and the classical geometry of de Sitter space, we consider de Sitter space both as a rigid background, on which
quantum fields propagate, and as a fluctuating spacetime in its own right. We review the wavefunction of an expanding universe, black holes in de Sitter space, and the de Sitter
entropy, and compare the cosmological horizon with the black hole horizon. The final three sections are dedicated to concrete models of de Sitter space in two, three and four spacetime dimensions.
Read the full abstract ▾
Machine Learning
Sharpness-Aware Minimization (SAM) Improves Classification Accuracy of Bacterial Raman Spectral Data Enabling Portable Diagnostics
Kaitlin Zareno, Jarett Dewbury, Siamak K. Sorooshyari et al. (+2 more)
Antimicrobial resistance is expected to claim 10 million lives per year by 2050, and resource-limited regions are most affected. Raman spectroscopy is a novel pathogen diagnostic approach promising rapid and portable antibiotic resistance testing within a few hours, compared to days when using gold standard methods. However, current
algorithms for Raman spectra analysis 1) are unable to generalize well on limited
datasets across diverse patient populations and 2) require increased complexity due to the necessity of non-trivial pre-processing steps, such as feature extraction, which are essential to mitigate the low-quality nature of Raman spectral data. In this work, we address these limitations using Sharpness-Aware Minimization (SAM) to enhance model generalization across a diverse array of hyperparameters in clinical bacterial isolate classification tasks. We demonstrate that SAM achieves accuracy improvements of up to 10.5% on a single split, and an increase in average accuracy of 2.7% across all splits in spectral classification tasks over the traditional optimizer, Adam. These results display the capability of SAM to advance the clinical application of AI-powered Raman spectroscopy tools.
Read the full abstract ▾
All authors Kaitlin Zareno, Jarett Dewbury, Siamak K. Sorooshyari, Hossein Mobahi, Loza F. Tadesse
Robotics
GLAMDRING: Gait Learning And Morphology co-Design via Reinforcement LearnING of CPGs
Amogh Joshi and Kaushik Roy
Robots are moving out of the structured factory floor and into unstructured environments such as disaster sites, planetary surfaces, and agricultural fields, for which the right
robot often does not yet exist. We present GLAMDRING, a framework that synthesizes the optimal
robot for a locomotion task and, jointly, learns the controller that drives it. For the given specifications of forward-velocity bounds, a per-actuator power budget, an actuator library, and a payload requirement, GLAMDRING returns a matched quadruped morphology (link geometry and per-joint actuators) and a
Hopf-oscillator Central Pattern Generator (CPG) gait policy. We rank feasible designs against a target design objective, viz., maximum speed, minimum Cost of Transport (CoT), or max Payload Margin. Because body and locomotion are coupled, the optimal morphology dictates how a
robot is driven, while optimal gait depends on the physical body. We train a small number of CPG policies by reinforcement learning across the space of candidate morphologies, co-learning the gait with the underlying
robot hardware. Link lengths and actuators are then resolved post-hoc from the policy's logged operating envelope, reducing synthesis cost to a small, fixed number of reinforcement-learning runs instead of one per candidate. Our experiments show three key findings: co-designing body and gait is necessary to satisfy locomotion constraints; actuator-envelope feasibility, rather than locomotion success alone, determines realizable payload capacity; and canonical animal gaits emerge naturally in most designs from morphology and constraints alone. A real-world demonstration further highlights the efficacy of our work.
Read the full abstract ▾
Computer Vision
Efficient Unified Multimodal Understanding (EUMU): Winning Solution for the MUMU Track at the 8th LSVOS Challenge
Dayoung Kil and Seong-heum Kim
The Mobile Unified
Multimodal Understanding (MUMU) Challenge requires a single efficient model to jointly perform multi-concept image tagging, open-vocabulary object detection, and image captioning. We present Efficient Unified
Multimodal Understanding (EUMU), the winning solution for the MUMU Track of the 8th LSVOS Challenge. EUMU builds on a shared pretrained
multimodal model, using its
prompt-based capabilities for detection and captioning and training lightweight heads on shared visual features to predict quality, scene, and event tags. Rather than treating the three tasks independently, EUMU applies task-aware inference refinement by reusing task outputs as cross-task cues. For detection, caption cues help recover objects missed by the initial detection. For captioning, detection cues help refine the caption to better reflect the detected objects. For tagging, image statistics refine quality predictions, while caption and detection cues refine scene and event predictions. This design unifies all three tasks within a single model while satisfying the challenge's resource constraints. EUMU contains 239.169M parameters, requires 23.947 GFLOPs, uses 4.5 GB of peak inference
memory, and achieves a final challenge score of 17.3409. Code and models are available at https://github.com/Dayoung-Kil/EUMU.
Read the full abstract ▾
Strongly Correlated Electrons
Energy partitioning of wave packets in one-dimensional systems
Flavia B. Ramos, Imke Schneider, Sebastian Eggert et al. (+1 more)
We investigate the non-
equilibrium dynamics of wave packets in a one-dimensional critical fermionic system and analyze the resulting partitioning of energy between emergent excitations. Starting from a Gaussian wave packet injected on top of the many-body ground state, we follow its real-time
evolution using the time-dependent density-
matrix renormalization group. We observe that interactions lead to fractionalization of the initial excitation into counter-propagating left- and right-moving modes, whose energies can be resolved in real space. To interpret these results, we employ Luttinger
liquid theory, which allows us to derive analytical predictions for the energy carried by the emergent modes. We find good agreement between field-theoretical predictions and numerical simulations in the low-energy regime. In contrast to charge fractionalization, which is completely determined by the Luttinger
liquid parameter, we show that energy partitioning is non-universal and depends on details of the injected wave packet, such as its width. Our results provide a real-space characterization of energy partitioning in one-dimensional systems and establish a quantitative comparison between non-
equilibrium numerical simulations and the effective field-theory description.
Read the full abstract ▾
All authors Flavia B. Ramos, Imke Schneider, Sebastian Eggert, Rodrigo G. Pereira
Robotics
Winning a Won Game: Strict Reach-Avoid-Stay Control Barrier Functions for High-Dimensional Black-Box Systems
Donggeon David Oh, Duy P. Nguyen, Gongkai Yuan et al. (+3 more)
Robots must complete their tasks and maintain the achieved outcomes while avoiding safety failures at all times. Strict reach-avoid-stay (sRAS) formalizes this requirement: safely reaching a target and remaining there indefinitely after first entry. We propose an sRAS Q-control barrier function (CBF) safety filter for high-dimensional black-box systems under bounded uncertainty. Our construction combines a stay value encoding safe permanent residence in a target subset with a reach-avoid value encoding safe reachability of this subset while avoiding target states from which safe permanent residence cannot be guaranteed. We prove that these values jointly yield a valid robust discrete-time CBF and lift them to state-action Q-functions for runtime intervention. For exact values and under a measure-zero condition, our filter preserves sRAS feasibility from almost every winnable initial state and keeps the system safely within the target after first entry, against all admissible uncertainty realizations. We adopt reachability-based
adversarial reinforcement learning for scalable value approximation using only black-box interactions. Notably, neither synthesis nor deployment of our filter requires known dynamics, affine structure, value derivatives, or hand-designed barriers. We validate our framework in quadruped gap jumping in simulation and hardware, where the
robot crosses the gap, lands safely, and remains safe afterward. Simulated F1TENTH races further demonstrate safe overtaking and lead retention.
Read the full abstract ▾
All authors Donggeon David Oh, Duy P. Nguyen, Gongkai Yuan, Qingchen Li, Jaime Fernández Fisac, Haimin Hu
Artificial Intelligence
The syntax and semantics of goals
David M. Abel and Mark K. Ho
In both cognitive science and computer science, goals are conceptualized as cognitive states that flexibly combine with world knowledge to organize and specify purposeful behavior. In this way, goals are compositional representations whose content relates to rational behavior. We here draw
attention to goals as representations and their content because it highlights a parallel with other areas in cognitive science - in particular, the syntax-semantics interface in linguistics and logic - while also foregrounding foundational questions about the expressivity, design, and efficiency of different goal representations. For example, goals are typically taken as fixed and imposing constraints on desirable behaviors, but we can also identify constraints on goal representations themselves, such as whether a particular goal
language is sufficiently expressive to capture behaviors of interest, or whether different goal representations capture the same behavior. Here, we synthesize work that aims to characterize the properties of different goal representations and suggest these are points of a broader design space. We close by discussing how distinguishing the form and meaning of goals can elucidate the implicit assumptions we make about goals, inform the study of interactions between higher-level cognition and motivation, and isolate axes of variation for different conceptions of goals.
Read the full abstract ▾
Robotics
From Wizard-of-Oz Human-Robot Dialogue Collection to a Taxonomy of Robot Response Decisions: A Retrospective Analysis of Assistive Pilot Interactions
Guangping Liu, Nicholas Hawkins, Tipu Sultan et al. (+2 more)
Robots that follow natural-
language instructions in everyday indoor environments must act on incomplete human utterances. Instructions often omit essential information, such as the identity of an out-of-view object, an intended destination, or the user's goal. Existing
datasets contain little real-world situated dialogue and provide few practice-grounded criteria for deciding when a
robot should act, confirm, clarify, or refuse. We retrospectively analyze a pilot Wizard-of-Oz study in which five participants performed everyday indoor tasks, including door opening, drawer opening, feeding, drinking, and cleaning, with a wheelchair-mounted mobile manipulator while the wizard responded without a formal communication policy. This preserved authentic user behavior but produced inconsistent
robot-side decisions, motivating an explicit decision scheme. From 40 episodes, we derived a hierarchical taxonomy of six response modes (ANSWER, REPORT_DONE, REFUSE, CONFIRM, CLARIFY, ACT) and four ambiguity types (intent, referential, spatial, intelligibility). Two human annotators and an AI annotator applied the scheme to the pilot data. Clean-label rates were 91% and 89%, and Cohen's ranged from 0.72 to 0.95 across decision-point, mode, and ambiguity levels for both human-human and human-AI comparisons. Fine-tuning LLaVA-1.6-7B on taxonomy-derived labels for ACT and CLARIFY indicates the feasibility of training
vision-
language models using annotations from our taxonomy. Remaining boundary cases in decision-point identification and REPORT_DONE motivate a constrained protocol for more consistent dialogue collection.
Read the full abstract ▾
All authors Guangping Liu, Nicholas Hawkins, Tipu Sultan, Flavio Esposito, Madi Dian
Algebraic Topology
$H\mathbb{Z}/4$ is not an $\mathbb{E}_2$-Thom Spectrum over the $2$-Complete Sphere Spectrum
Mattie Ji
For any
prime number $p$, the classic Hopkins-Mahowald
theorem asserts that $H\mathbb{Z}/p$ is an $\mathbb{E}_2$-Thom spectrum over the $p$-complete sphere spectrum. Kitchloo extended this result to show that $H\mathbb{Z}/p^k$ is an $\mathbb{E}_2$-Thom spectrum over the $p$-local sphere spectrum, except for the case when $(p, k) = (2, 2)$. We solve the remaining case by showing that $H\mathbb{Z}/4$ is not an $\mathbb{E}_2$-Thom spectrum over the $2$-complete sphere spectrum, and consequently the $2$-local sphere spectrum. This work was developed with substantial assistance from GPT-5.6 Sol and GPT-6 Astra, and an early draft was subsequently reviewed using Claude Fable 5.1.
Read the full abstract ▾
Multimedia
From Models to Systems: A Comprehensive Survey of Efficient Multimodal Learning
Pan Wang, Siwei Song, Hui Ji et al. (+12 more)
The rapid expansion of
multimodal models has surfaced formidable bottlenecks in computation,
memory, and deployment, catalyzing the rise of Efficient
Multimodal Learning (EML) as a pivotal research frontier. Despite intensive progress, a cohesive understanding of what, how, and where efficiency is manifested across the learning stack remains fragmented. This survey systematizes the EML landscape by introducing the first structured, model-to-system taxonomy. We distill insights from over 300 seminal works into three hierarchical levels–model,
algorithm, and system–addressing architectural parsimony, execution refinement, and hardware-aware orchestration, respectively. Moving beyond a purely categorical review, we offer a methodological synthesis of the vertical synergies between these layers, elucidating how cross-layer co-design contributes to the fundamental "Efficiency-Utility-
Privacy" trade-off. Through an integrative case study of
Multimodal Large
Language Models (MLLMs), we trace the field's evolutionary trajectory from initial structural adjustments to modern full-stack resource orchestration. Furthermore, we provide a holistic discussion and application-specific optimization blueprints for diverse domains and posit a paradigm shift toward self-regulating intelligence, where efficiency is an intrinsic, emergent property of the model's fundamental design rather than a post-hoc constraint. Finally, we present open challenges and future directions that will define the trajectory of EML research. This survey establishes a structured framework for
multimodal systems that are not only high-performing and generalizable but natively efficient and ready for ubiquitous deployment. A continuously updated version is available at https://github.com/pwang322/Efficient-
Multimodal-Learning-Survey.
Read the full abstract ▾
All authors Pan Wang, Siwei Song, Hui Ji, Siqi Cao, Heng Yu, Zhijian Liu, Huanrui Yang, Yingyan Celine Lin, Beidi Chen, Mohit Bansal, Xiaoming Liu, Pengfei Zhou, Ming-Hsuan Yang, Tianlong Chen, Jingtong Hu
Computer Vision
Seeing Abnormal from Normal: Glomerular Abnormality in Representations of Normal Renal Morphology
Greta Hasko, Rachit Saluja, Tianyu Shi et al. (+11 more)
Fine-grained evaluation of glomerular pathology must distinguish normal glomeruli from abnormalities such as global and segmental glomerulosclerosis, obsolescent, ischemic, solidified, disappearing, and atubular glomeruli. Supervised classification requires labeled examples of every category, which is impractical when subtypes are rare or absent from the training cohort. One-class anomaly detection offers an alternative by modeling normal data and scoring deviations, allowing previously unseen abnormalities to be detected. We use the frozen residual U-Net backbone of Omni-Seg, pretrained to segment structurally normal renal primitives without abnormal-subtype labels. We propose NoRDeC (Normal-Reference Detection and Characterization), a framework combining Mahalanobis normal-reference scoring with layer-wise representation analysis to determine whether and where glomerular pathology is encoded, how spatial aggregation affects detection, and whether abnormalities alter inter-layer relationships differently. Using glomerular images from two institutions, we evaluate backbone layers and aggregation strategies, compare NoRDeC with PaDiM and PatchCore, and analyze representations using centered kernel
alignment (CKA). Layer 4 with Center-70 aggregation achieved a pooled AUROC of $0.926\pm0.013$. NoRDeC achieved the highest AUROC in six of seven abnormality categories and in the pooled analysis, while CKA suggested subtype-dependent changes in inter-layer relationships not captured by anomaly scores alone. The normal-reference model is fitted using only normal glomeruli; abnormality labels are used for configuration selection, evaluation, and grouping in the representation analysis. These results show that a frozen renal feature extractor can support both detection and representation-level characterization of glomerular abnormalities without using abnormal examples to fit the detector.
Read the full abstract ▾
All authors Greta Hasko, Rachit Saluja, Tianyu Shi, Leiyue Zhao, Yuechen Yang, Daniel Reisenbuechler, Tianyuan Yao, Zhenhao Guo, John Cannon, Yuling Chi, Lorraine Gudas, Mert R. Sabuncu, Yihe Yang, Ruining Deng
Statistics Theory
Robust Error Bounds for Vector-Valued Kernel Ridge Regression
Wenbin Wang and Colin N. Jones
While kernel-based learning methods for scalar-valued functions have been extensively studied, comparatively little work considers nonlinear vector-valued operators. Such operators with vector-valued outputs, e.g., state vectors and trajectories, commonly arise in system identification problems. In this paper, we study the problem of learning a vector-valued operator under the framework of vector-valued kernel ridge regression. We assume that the outputs take values in an arbitrary
Hilbert space and are corrupted by additive noise with bounded norm. Under the assumption that the unknown operator belongs to a vector-valued reproducing kernel
Hilbert space, we derive a pointwise deterministic bound on the discrepancy between the kernel ridge regression
estimator and the ground truth in the corresponding
Hilbert space norm. Numerical simulations under two different scenarios are presented to validate the theoretical results.
Read the full abstract ▾
Mathematical Physics
Site-scale, Multi-resolution WRF Wind Modeling with Observational Nudging for Methane Emission Monitoring in the Permian Basin
Ehsan Kharazmi, Erin Menezes, George Karniadakis et al. (+2 more)
Accurate methane source localization and emission-rate estimation at oil and gas facilities require wind fields that resolve site-scale spatial and temporal variability, which cannot be characterized by a single anemometer or coarse operational weather products. We develop a multiscale Weather Research and
Forecasting (WRF) framework for a Permian Basin facility that dynamically downscales hourly, 3 km High-Resolution Rapid Refresh (HRRR) fields through four nested domains with grid spacings of 3 km, 1 km, 200 m, and 40 m. The two outer domains use planetary-boundary-layer parameterization, whereas the two inner domains operate in large-eddy-simulation mode. One-minute wind observations from two on-site anemometers are assimilated through wind-only observational nudging in the innermost domain, and the effects of nesting feedback and
turbulence-closure choices are examined. Simulations for winter and summer 2025 periods are evaluated using near-surface wind-speed time series, power spectra, spatial fields, and statistical metrics. The baseline simulation reproduces the broad
evolution of observed wind events but drifts during weak-wind periods. Observational nudging with one- and two-way nesting reduces the absolute bias of wind-speed by %65 at one sensor and %87 at the other, while reducing the corresponding root-mean-square errors by %33 and %44. Nudging also increases high-frequency energy and resolved spatial gradients, although improvements in
turbulence statistics are not uniform. These results demonstrate a practical physics-based pathway for generating high-resolution wind fields for methane-plume modeling while emphasizing the need for independent spatial observations to validate accuracy away from assimilated sensors.
Read the full abstract ▾
All authors Ehsan Kharazmi, Erin Menezes, George Karniadakis, Anantha Sundaram, Arash Fathi
Robotics
Predict Before You Deploy: Offline Prediction of Quantization-Induced Task Degradation for World Action Models
Jiuyi Xu, Jinjia Guo, Meida Chen et al. (+2 more)
World action models (WAMs) rely on
video-generation backbones, requiring substantial
memory and compute for deployment. Post-training quantization reduces
memory and can accelerate inference, but bit width, grouping, and quantizer choice define a large configuration space. Identifying configurations that preserve task performance through exhaustive closed-loop evaluation is costly. We propose PreDE (Predict Before You Deploy), a policy-calibrated framework for predicting quantization-induced task degradation from offline action deviations. Using closed-loop outcomes from a small development set, PreDE calibrates two thresholds and accepts, rejects, or defers new configurations using a fixed observation log. Under a within-setting label-ordering hypothesis, the rule issues decisions where all thresholds consistent with the development labels agree. Across five WAMs and four
benchmark settings, quantization produces configuration-dependent task losses that cannot be explained by bit width alone or a shared deviation threshold. Across 28 held-out configurations from two policies, PreDE issued 21 decisions before observing closed-loop outcomes (75% coverage), all matching the observed acceptable or degraded labels. Deferred candidates included both acceptable outcomes and a 33-percentage-point loss. In 450 Franka Research 3 trials across two independently fine-tuned policies, all configurations assigned to high-deviation groups before testing showed significant degradation, while low-deviation comparisons showed no statistically significant degradation. On the real
robot, W4A4 achieved a 1.37x action-query speedup and approximately 44% lower peak
memory. These results support policy-specific behavioral calibration for quantization configuration selection while identifying candidates that require closed-loop evaluation. The code is available at https://github.com/jiuyixu25/PreDE.
Read the full abstract ▾
All authors Jiuyi Xu, Jinjia Guo, Meida Chen, Jing Du, Yangming Shi
Number Theory
A uniform proof approach for congruences modulo 3 for partitions with $k$-colored odd parts
Marie Jameson and James A. Sellers
In recent work, Hirschhorn and the second author defined $a_k(n)$ to be the number of partitions of $n$ wherein the even parts come in only one color, while the odd parts may be “colored” with one of $k$ colors for fixed $k\geq 1$. This function generalizes the classical partition function and has been of significant interest because it satisfies a number of congruences. Although prior work studying $a_k(n)$ has resulted in congruences in arithmetic progressions in a somewhat ad hoc manner, this work gives a uniform framework for studying congruences modulo 3. This allows us to prove an infinite family of infinite families of non-nested congruences modulo 3.
Read the full abstract ▾
Number Theory
Endoscopic description of the local Langlands correspondence for $\mathrm{G}_2$
Yugo Takanashi
We prove that the $L$-packets for $p$-adic $\mathrm{G}_2$ constructed by Gan and Savin satisfy the endoscopic character identities. In the course of the proof, we also prove twisted endoscopic character identities for functorial lifts from $\mathrm{G}_2$ to $\mathrm{PGSO}_8$ with triality. As a byproduct, we also prove a global multiplicity formula for the discrete automorphic spectrum of $\mathrm{G}_2$ coming from the cuspidal automorphic spectrum of $\mathrm{PGL}_3$.
Read the full abstract ▾
Game Theory
On Periodic and Aperiodic Optimal Strategies in Solvency Games
Quentin Guilmant, Florian Luca, Richard Mayr et al. (+2 more)
Solvency games are a gambling problem on infinite-state Markov decision processes in which the state $n \in \mathbb{N}$ represents an investor's fortune. In every round, the investor chooses an action from a finite action set, and every action yields a distribution over integer-valued gains in an interval $\{-\ell,\ldots,m\}$. The risk-averse investor wants to minimise the probability of eventual ruin (reaching a fortune $\le 0$). It was shown in [Berger et al.] that memoryless deterministic optimal strategies exist, but they are not eventually constant in general. Even in the special case of gains in $\{-2,\ldots,1\}$, the optimal strategy may need to make use of two different actions at arbitrarily high fortunes. We show that optimal strategies in solvency games need not be ultimately periodic in general (thus disproving a 2012
conjecture of Kučera). Already in the case of gains in $\{-3,\ldots,1\}$, it is possible for the optimal strategy to be unique but aperiodic. For gains in $\{-2,\ldots,1\}$, there always exists an ultimately periodic optimal strategy whose tail is constant or alternates between two actions. Finally, we show that the optimal strategy is computable if it is unique. Moreover, (some) optimal strategy can always be computed in the case of gains in $\{-\ell,\ldots,1\}$ for any $\ell \in \mathbb{N}$. Computability in the general case however remains open.
Read the full abstract ▾
All authors Quentin Guilmant, Florian Luca, Richard Mayr, Joël Ouaknine, James Worrell
Machine Learning
Bayesian Optimization with Rich Auxiliary Information via LLMs
Efe Mert Karagözlü, Rohit Sonker, Tejus Gupta et al. (+2 more)
Bayesian Optimization (BO) is widely used for optimizing expensive black-box functions, yet it typically reduces each expensive experiment to an input and its objective value, discarding much of the information the experiment produces. Such auxiliary information can include training curves in hyperparameter optimization, expert notes and images in scientific experimentation, or known facts about the specific optimization problem. Leveraging the ability of large
language models (LLMs) to process diverse and unstructured information, we introduce two methods: AuxBO-Evolve, which uses auxiliary information to evolve beliefs over the location of the optimum, and AuxBO-Vicinity, which uses it to locally guide the acquisition function. Across synthetic tasks, hyperparameter optimization
benchmarks, and a
nuclear fusion task with unstructured scientist-written logs, our methods consistently improve optimization performance and outperform standard BO and existing LLM-based approaches. We further find that LLMs provide more effective probabilistic guidance when modeling likely maximizer locations rather than objective values pointwise. Overall, our results demonstrate that exploiting rich information beyond standard $(\mathbf{x}, y)$ interactions can substantially improve BO performance.
Read the full abstract ▾
All authors Efe Mert Karagözlü, Rohit Sonker, Tejus Gupta, Barnabás Póczos, Jeff Schnieder
Pattern Formation and Solitons
The Origin of Imperfection Sensitivity in the Buckling of Cylindrical Shells
Tian Yang and Tobias M. Schneider
Buckling of thin cylindrical shells under axial compression, a classical example of a subcritical instability, is highly sensitive to small imperfections, with minute
geometric variations causing large changes in buckling threshold. To uncover the origin of this sensitivity, we use numerical continuation and bifurcation analysis while systematically varying the depth and size of a single localized Gaussian defect. We show that the instabilities of the imperfect shell originate from localized
equilibria already present in the perfect shell. By breaking translation
symmetry, the defect pins these
equilibria and changes how they connect to the imperfect base state. Small changes in defect geometry can thereby switch the bifurcation that triggers buckling, producing non-monotonic and discontinuous changes in buckling threshold and abrupt changes in buckling mode. Imperfection sensitivity is therefore not simply sensitivity to imperfection magnitude, but sensitivity of the underlying bifurcation structure to imperfection geometry.
Read the full abstract ▾