Nuclear Theory
Levy-shaped femtoscopic sources from an exact solution of relativistic hydrodynamics
Bence Szentkuti and Mate Csanad
An important goal of high-energy physics is to better understand the spacetime structure of particle emitting sources created in heavy-ion collisions. For a fraction of a second a special form of matter, the so-called
quark-
gluon plasma (QGP) is created, which can be examined with the toolkit of femtoscopy. With the measurement of correlation functions it was determined that Lévy distributions describe the shape of QGP sources the most accurately, however, based on hydrodynamics Gaussian distributions are typically expected. In this article we showcase an exact solution of the equations of relativistic hydrodynamics, which aims to resolve this puzzle by providing Lévy distributions directly from hydrodynamics. A detailed analysis reveals whether the predictions of this solution are in accordance with observations.
Read the full abstract ▾
Social and Information Networks
Polarization Dynamics under Algorithmic Recommendations in LLM-based Online Social Networks
Ali Safarpoor Dehkordi, Mohammad Shirzadi, Mehwish Nasim et al. (+1 more)
How do recommendation
algorithms contribute to the formation of polarized opinions in online social networks? This question has been studied through empirical analyses of real platforms and mathematical models of opinion dynamics. Empirical studies capture user behavior but are often limited by noisy, incomplete, or inaccessible platform data, while mathematical models offer control and interpretability but usually rely on simplified opinion update rules. More recently, large
language model (LLM)-based social simulations have provided an alternative for studying recommendation and opinion dynamics through natural-
language interactions. However, how different ways of applying recommendations shape opinion dynamics remains underexplored. In this work, we provide a controlled comparison of two different recommendation mechanisms in an LLM-based online social network simulation. Edge Recommendation (ER) introduces new connections in the network, while Content-Exposure Recommendation (CR) exposes additional users to individual posts without modifying the underlying graph. Our framework represents users as LLM-based
agents who interact in natural
language while maintaining evolving opinion states. We measure polarization as opinion dispersion, extremization as the average distance of opinions from the neutral midpoint, and reinforcement as interactions that amplify already aligned opinions. Our results show higher mean polarization and extremization across recommendation conditions compared with the no-recommendation baseline, with absolute differences of up to $0.12$ in our main experiments.
Read the full abstract ▾
All authors Ali Safarpoor Dehkordi, Mohammad Shirzadi, Mehwish Nasim, Ahad N. Zehmakan
Machine Learning
Hidden Data Challenges in Clinical Machine Learning
Carolin Cissée, Dennis Fast, Raneen Younis et al. (+3 more)
Widely reused preprocessing pipelines for electronic health records can encode assumptions about when information became available, what clinical episode a label represents, and what missingness means. These assumptions are rarely audited systematically and may propagate across subsequent studies when preprocessing pipelines are reused or adapted. Using MIMIC-IV as a case study, we identify and quantify violations of three complementary forms of preprocessing
alignment, i.e., temporal, semantic, and population
alignment, which manifest as temporal leakage, feature-task misalignment, and biased imputation. We show how ambiguities in timestamp semantics, episode definitions, endpoint construction, modality selection, and missingness handling can alter the information available to a model and affect the interpretation of its predictions. Through analyses of MIMIC-IV data and empirical case studies, we demonstrate that such choices may remain largely invisible in aggregate predictive performance while changing the signals on which models rely. We derive practical recommendations for temporally valid, clinically aligned, and reproducible preprocessing of MIMIC-IV and related electronic health record
datasets. The code for our experiments, including our proposed preprocessing, is available at: https://github.com/m2l-lab/mimic-iv-preprocess.
Read the full abstract ▾
All authors Carolin Cissée, Dennis Fast, Raneen Younis, Alexander Löser, Wolfgang Nejdl, Zahra Ahmadi
Quantum Physics
A General Method for Extending Isoscattering Non-Isospectral Quantum Graphs
Omer Farooq, Michal Lawniczak and Leszek Sirko
The present paper sets out a general method for extending the isoscattering, non-isospectral
quantum graphs. The methodology is illustrated by considering two graphs, denoted by $\Gamma_1$ and $\Gamma_2$, which have the same total metric length but different
topologies. It is demonstrated that, despite the fact that their scattering
matrices are identical, their
eigenvalues, denoted by $k_n$, and Euler characteristics differ from one another. The next step in the process is to extend both graphs by affixing the same
star graph to each of them. Utilising the Titchmarsh-Weyl M-function formalism, it is proven that the newly constructed, extended graphs are also isoscattering and non-isospectral.
Read the full abstract ▾
Computation and Language
Test-Time Memory Construction
Zijun Wang, Yifan Ji, Haidong Xin et al. (+6 more)
Large
Language Models (LLMs) struggle to answer questions over long interaction histories, where relevant evidence may be scattered across distant sessions. Existing
memory systems typically construct and maintain sophisticated
memory representations offline, before user queries are observed, incurring substantial overhead and limiting their ability to adapt
memory construction to query-specific information needs. In this paper, we propose T$^2$Mem, a Test-Time
Memory construction framework that defers
memory construction to inference and progressively builds
memory along the
reasoning process. Specifically, T$^2$Mem extracts lightweight episode-level metadata offline to capture key events and temporal associations as
retrieval cues, while retaining links to the original interactions to facilitate the reconstruction of dependencies across episodes. At test time, the LLM adaptively retrieves and organizes metadata and raw interactions, progressively constructing query-specific
memory as
reasoning unfolds. Experiments on LoCoMo and LongMemEval show the effectiveness of T$^2$Mem by consistently outperforming strong baselines across different backbone LLMs while substantially reducing the overhead of
memory construction. Further analysis demonstrates that deferring
memory construction to inference enables the
memory construction process to dynamically adapt to the information needs of individual queries. All code is available at https://github.com/OpenBMB/T2Mem.
Read the full abstract ▾
All authors Zijun Wang, Yifan Ji, Haidong Xin, Zhenghao Liu, Yukun Yan, Zulong Chen, Yu Gu, Ge Yu, Maosong Sun
Geometric Topology
Thurston's Legacy at the Interface of Surface Dynamics, Train Tracks, and Arithmetic
Ahmad Rafiqi and Ayberk Zeytin
In this chapter, we discuss the interaction between low-dimensional dynamics, symbolic dynamics, and
geometric topology, focusing on the algebraic characterization and asymptotic behaviour of stretch factors or equivalently pseudo-Anosov dilatations. We first review the algebraic foundations linking Perron–Frobenius theory, multi-variable Euclidean
algorithms, and Lind's characterization of topological
entropy for shifts of finite type, emphasizing recent constraints on the Perron–Frobenius degree of expansion constants. We then relate these dynamical invariants to the combinatorial world, examining train tracks for surfaces and
automorphisms of free groups. Finally, we briefly review the relevance of fibered 3-
manifolds, detailing how the Thurston norm, Fried's concavity
theorem over fibered cones, and McMullen's Teichmüller
polynomial come into play.
Read the full abstract ▾
Strongly Correlated Electrons
Interplay of disorder and spin liquid physics in the triangular-lattice antiferromagnet YbZn2GaO5
Hafsa Zeroual, Shams Sohel Islam, Toni Shiroka et al. (+7 more)
YbZn2GaO5 has recently emerged as a structurally ideal triangular-
lattice antiferromagnet and a promising candidate for hosting a U(1) Dirac
quantum spin-
liquid (QSL) ground state. However, signatures of broadened
crystal electric field (CEF) excitations have raised questions about potential site disorder and its impact on the intrinsic
spin dynamics. In this study, we employ 71Ga
nuclear magnetic resonance (NMR) and
muon spin relaxation (muSR) to microscopically probe both the structural and
magnetic properties of YbZn2GaO5. NMR spectra reveal two inequivalent 71Ga sites and a broad distribution of quadrupolar frequencies, providing clear evidence of Ga$^{3+}$/Zn$^{2+}$ site mixing and local CEF randomness. The NMR
spin-
lattice relaxation rate exhibits, down to 1.3~K, a power-law temperature dependence, consistent with persisting
spin dynamics below the exchange energy scale in a frustrated magnet. Complementary muSR measurements further show that the
spin dynamics persist down to 30mK, with no signs of
spin freezing. Together, these results establish that YbZn2GaO5 presents substantial structural disorder, while the persistence of dynamic
spin fluctuations down to the lowest temperatures may indicate the survival of QSL-like excitations despite such disorder.
Read the full abstract ▾
All authors Hafsa Zeroual, Shams Sohel Islam, Toni Shiroka, Thomas James Hicken, Rabindranath Bag, Sara Haravifard, Hubertus Luetkens, Philippe Mendels, Edwin Kermarrec, Fabrice Bert
Robotics
Evo-RL: Iterative Robot Reinforcement Learning from Selective Experience
Kai Ye, Yifan Jin, Qianhang Luo et al. (+7 more)
Existing real-world reinforcement learning methods improve
robot policies by combining human corrections with autonomous rollouts. However, accumulated experience varies in its value for policy improvement, and even successful trajectories can contain inefficient or redundant behavior, motivating more selective experience reuse. To this end, we propose Evo-RL, an iterative reinforcement learning method that combines human interventions with advantage-guided experience selection. Evo-RL uses human corrections and experience with positive estimated advantages from successful trajectories to iteratively refine the policy. To support learning from these different experience sources, we introduce RW-RL-
Dataset, which spans over 1,000 hours across four
robot embodiments. On a real-world high-precision manipulation task, Evo-RL improves attempt success rate over RECAP by 30.46%. Our code, model checkpoints, and
dataset are publicly available.
Read the full abstract ▾
All authors Kai Ye, Yifan Jin, Qianhang Luo, Hongru Zhang, Jiedong Zhang, Shuyuan Hu, Song-Yuan Wu, Jihai Xu, Jie Zhao, Bo Zhao
Computer Vision
Why DiT Models Underperform as Representation Learners without Long Skip Connections
Benyuan Meng, Yiliang Zhang, Jin-Wen Wu et al. (+3 more)
Diffusion feature, i.e., intermediate activations extracted from
diffusion model backbones, has emerged as a promising approach for unifying generation and
vision understanding (discrimination). Despite the advancement of
diffusion backbones into
Diffusion Transformers (DiT), it has been found in previous literature that DiT models tend to yield weaker features despite their stronger generation capability. This poses a threat to good generation-understanding unification. In this work, we show via controlled from-scratch training that long skip connections (LSCs) affect feature quality: a DiT with LSCs can produce better features. Furthermore, through detailed diagnosis, we reveal the mechanism causing this effect. Specifically, the input to a
diffusion backbone contains both the latent image and the
diffusion noise, where the
diffusion noise, though appearing in the form of noises, actually reflects the information of the
diffusion trajectory. To make the final prediction of a noisy variable ($v$ or $\epsilon$), the
diffusion backbone needs both signals. However, carrying both signals through the main backbone effectively "pollutes" the image signal with noises, leading to poor feature quality. In comparison, LSCs can shortcut
diffusion noises directly to where they are needed, leaving much weaker noise signals in the main backbone, thus improving feature quality. To name the broader impact, this reveals that LSCs are not a mere trick but closely related to the
diffusion objective. The code is at https://github.com/Darkbblue/dit-lsc.
Read the full abstract ▾
All authors Benyuan Meng, Yiliang Zhang, Jin-Wen Wu, Qianqian Xu, Qingming Huang, Longtao Huang
Analysis of PDEs
Loss of derivatives for hyperbolic operators with coefficients oscillating in time
Wenhui Chen and Michael Reissig
We consider second-order strictly
hyperbolic operators whose coefficients are Lipschitz continuous in space and may oscillate near the initial time. Their first two time derivatives are bounded by the corresponding powers of $\frac{1}{t}(\log\frac{1}{t})^\alpha$ with $0\leqslant\alpha<1$, and the first time derivative satisfies the corresponding spatial Lipschitz bound. By combining Tarama's energy with frequency-dependent weights, we prove an energy estimate without loss for $\alpha=0$ and with a stretched logarithmic upper bound for the loss when $0<\alpha<1$, which implies an arbitrarily small loss in the Sobolev scale. These estimates extend the analysis of Colombini, Del Santo and Fanelli [Ann. Scuola Norm. Sup. Pisa Cl. Sci. (5) 26 (2025)] for spatially Lipschitz coefficients to time-derivative bounds with an additional logarithmic factor.
Read the full abstract ▾
Mesoscale and Nanoscale Physics
Atomic Edge Resonators Flatten Phonon Bands in Biphenylene Nanoribbons
V. L. Katkov, N. M. Chtchelkatchev and V. A. Osipov
We show that H–F edge complexes in biphenylene nanoribbons act as intrinsic atomic resonators, flattening low-frequency
phonon branches. With increasing wave vector, flexural and twisting vibrations concentrate at the edges while motion of the carbon backbone is suppressed. Displacement profiles and edge-mass tuning identify a local-resonance mechanism, an atomic counterpart of a tuned mass damper. Combining machine-learned interatomic potentials, frozen-
phonon calculations, and density-functional tight binding, we find that weakly dispersive modes contribute substantially to thermal band-gap narrowing. Zero-point motion shifts the calculated gap closure toward the experimentally observed ribbon widths. A hexagonal boron nitride substrate preserves the principal spectral features while reducing thermal renormalization. These results establish edge chemistry as a route to nearly flat
phonon bands and connect a local atomic resonance to the electronic response of an entire nanoribbon.
Read the full abstract ▾
Machine Learning
Understanding Mamba-3 Through Oscillatory Dynamics
Shutian Liang and Yue Song
Mamba-3 introduces input-dependent rotations into recurrent state updates, but their contribution beyond state retention remains unclear. We interpret these rotations as underdamped second-order dynamics, providing a framework for investigating the role of oscillation in sequence modeling. Controlled comparisons show that underdamped dynamics outperform non-oscillatory damping regimes on state-tracking tasks, while
language-model pretraining demonstrates benefits for long-context prediction and length generalization. Interventions in pretrained models further establish the importance of learned phase, and state analysis suggests that rotation regulates how retained information combines with incoming updates. Taken together, these findings identify underdamped dynamics as a useful mechanism for recurrent computation and suggest that oscillation supports long-context processing by regulating how information accumulates.
Read the full abstract ▾
Computers and Society
Who Can Say No? Control, Accountability and the Pace of Frontier AI
Rikard Rosenbacke, Carl Rosenbacke, Victor Rosenbacke et al. (+1 more)
In September 2026, the leaders of the largest frontier AI laboratories said in public that the pace of capability improvement is too fast and that none of them can slow down alone, and
agents under evaluation at OpenAI broke out of their test environment and reached a third party's live systems, which a UN-mandated panel called an early warning of one route to loss of human control. This paper examines the institutional side of that risk: not the machine, but the person who must control it. We argue that control is not taken by AI but given away, one delegation at a time, by organisations arranged so that nobody is positioned to say no, and we define control as a person who can detect a serious problem, challenge the decision, stop the system and reverse it, and who has something to lose by not doing so. Three questions follow, which any builder, funder, regulator or deployer can answer in public: (1) who inside the organisation can say no, and who answers for saying yes; (2) who outside the organisation checks the decision, and what can they do when they disagree; (3) what happens when AI builds AI faster than anyone can check it. A checkable answer needs a named person, a date, a number, and a stop that has been used or tested. Three measures meet that test: a named person on every consequential decision, including the standing instructions AI
agents carry out; a supervisor with a stop authority; and two published clocks, time to build against time to check. We show that governance theory arrives at the same conditions, and code fifty older mechanisms against the questions, finding that almost every one began as one rule at one point, taken by one party without waiting for the rest. That is the paper's design: a start, not a solution. Publication is the first step and one laboratory can take it alone; the contract, the independent body and the law come after.
Read the full abstract ▾
All authors Rikard Rosenbacke, Carl Rosenbacke, Victor Rosenbacke, Martin McKee
Combinatorics
Completely mobile robots in graphs with a resolving property
Jose C. Camacho Moreno, Dorota Kuziak, Ethan Shallcross et al. (+2 more)
This work extends the study of mobile resolvability in connected graphs by introducing completely mobile resolving sets. As with mobile resolving sets, we consider stationing a collection of
robots at the vertices in a resolving set of a graph.
Robots may move one at a time to any adjacent vertex as long as set of occupied vertices remains a resolving set. In this new variant, we ask that every
robot can traverse every vertex in the graph through a sequence of permitted moves. Graphs possessing at least one completely mobile resolving set are termed mobile-resolvable. We completely characterise the non-mobile-resolvable graphs as paths of order at least three and subdivided
stars with three or more leaves and at most two subdivided edges. For trees, we explicitly determine the completely mobile metric dimension, showing that it exceeds the classical metric dimension by one for most mobile-resolvable trees, and by two for certain subdivided
stars. Furthermore, we evaluate this parameter across fundamental graph classes-including cycles, complete bipartite graphs, grid graphs, and circulant graphs.
Read the full abstract ▾
All authors Jose C. Camacho Moreno, Dorota Kuziak, Ethan Shallcross, James Tuite, Ismael G. Yero
Computer Vision
VideoAWR: Reinforcing Video Flow Models with Decoupled-Timestep Advantage-Weighted Rectification
Zheming Fu, Wenbin Ning, Ruizhe He et al. (+6 more)
Regression-based RL for flow models mandates sampling without classifier-free guidance (CFG) to align the rectification target with the reference marginal velocity. Extending Advantage-Weighted Rectification (AWR), which dynamically scales this regression target based on reward, to
video generation encounters a cold-start pathology: unguided rollouts collapse into low-fidelity regions where intra-group reward gaps remain wide but no longer track perceptual quality, rendering the optimization target arbitrary. Conversely, utilizing guided rollouts to restore fidelity, while keeping the regression target tied to the guided sampling distribution, provably triggers a
geometric guidance blow-up under exponential-moving-average (EMA) updates. We propose VideoAWR, which resolves this tension by untying sampling from the optimization anchor. An EMA reference policy generates guided rollouts to restore reward separability, while the trainable policy is rectified against the unguided conditional anchor. We prove this decoupled objective is mathematically exact: the centered advantage cancels the anchor in expectation, filtering out the reward-neutral guidance offset and retaining only the reward-correlated covariance increment. Consequently, inference runs single-branch and CFG-free, yet inference-time CFG preserves its original semantics. Furthermore, we introduce Decoupled Timestep Rectification, supervising each rollout at an independently sampled timestep without bias to improve sample efficiency. Experiments on Wan2.1-1.3B and 14B demonstrate stable optimization and superior visual and motion quality in both CFG-free and CFG-enabled regimes. Code is available at \pdfstartlink attr{/Border [0 0 0]} user{/Subtype /Link /A << /S /URI /URI (https://github.com/Fuzm7/FlowAWR) >>}% \textcolor{pearDark}{VideoAWR}\pdfendlink.
Read the full abstract ▾
All authors Zheming Fu, Wenbin Ning, Ruizhe He, Wei Shang, Xiaoxiao Ma, Lianrui Mu, Chang Liu, Lei Wang, Siming Fu
Computer Vision
Exploring Attention Sinks in Text-to-Image Transformers from Diffusion and Autoregressive Models
Binglei Li, Mengping Yang, Zhiyu Tan et al. (+3 more)
Attention sinks are
tokens or positions that consistently receive disproportionate
attention mass, often beyond what their semantic content would suggest. Despite extensive study in large
language models, their behavior and functional relevance in text-to-image (T2I)
Transformers remain largely unexplored. To fill this gap and better understand how text-to-image
Transformers allocate and route
attention, we explore this phenomenon of text-to-image
transformers across bidirectional
diffusion models and causal autoregressive models. Specifically, we explore whether
attention sinks exist, where they emerge, and how they functionally influence generation. Through systematic
attention analysis and targeted interventions, we identify that
attention sinks emerge in both bidirectional
diffusion and autoregressive models but exhibit distinct patterns. The autoregressive model concentrates
attention on the first and global-scale
tokens, whereas the bidirectional models develop sinks near modality boundaries, within text regions, and at spatial corners. Moreover, these patterns vary across generation stages and are localized to a sparse subset of layers and heads. To assess their functional relevance, we perturb sink-related heads and
token positions. It turns out that perturbing sink-ranked heads produces larger average changes than non-sink comparisons, while some sink-position perturbations have limited effects on bidirectional model scores. We further introduce register
tokens to redistribute
attention sinks and find that several configurations preserve or modestly improve generation performance. Overall,
attention sinks are architecture-, inference-stage-, and component-dependent phenomena, providing a basis for mechanistic analysis and principled
attention design in text-to-image
transformers.
Read the full abstract ▾
All authors Binglei Li, Mengping Yang, Zhiyu Tan, Wenxu Qian, Junping Zhang, Hao Li
Signal Processing
Scalable Quantum Clustering of Renewable Profiles via a Hybrid Qubit-efficient Strategy
Yifeng Cao, Chengjun Liu, Yijun Xu et al. (+4 more)
Large-scale renewable energy profile clustering is a computationally challenging NP-hard problem. Recent
quantum computing techniques offer a potential remedy, yet their scalability is severely constrained by the limited
qubit count of current
quantum hardware. To address this issue, we first cast the problem into a QUBO form and then we propose a consensus-based variable reduction (CBVR) method to drastically cut
qubit usage. More specifically, we tailor the one-hot constraint to eliminate assignment variables and leverage consensus analysis to convert stable clustering relations from initial anneals into equality constraints for further reduction. Experiments on real
quantum hardware confirm that CBVR compresses oversized QUBO models into feasible sizes while retaining high clustering quality and efficiency.
Read the full abstract ▾
All authors Yifeng Cao, Chengjun Liu, Yijun Xu, Kai Wen, Wei Gu, Lamine Mili, Shuai Lu
Machine Learning
AgentTune: A Unified Training Lifecycle for Tool-Using LLM Agents
R E Zera Marveen Lyngkhoi, Abhivansh Gupta, Vidushee Vats et al. (+3 more)
Tool-using
agents increasingly require a common substrate for building, training, evaluating, distilling, and maintaining agentic systems, yet existing tooling typically addresses these stages in isolation. We present AgentTune, a source-available Python library that provides a unified framework for developing and optimizing tool-using
agents. At its core, AgentTune introduces a two-tier EventLog representation that preserves
agent trajectories while supporting consistent projections for training, evaluation, and
auditing. Building on this substrate, the library provides a shared
agent runtime, tool-calling reinforcement learning, agentic
distillation,
retrieval-augmented generation, decision workflows, evaluation utilities, and a designed path toward self-healing
agents. These components are exposed through composable interfaces so that
agent strategies, training methods, evaluation procedures, and recovery mechanisms can be combined without requiring separate representations or pipelines. We demonstrate the framework across agentic tasks including legal contract search and domain-specific applications, illustrating its use for trajectory-based optimization, tool-use evaluation, and cross-task
agent development. AgentTune is released as a source-available library at https://github.com/Lexsi-Labs/AgentTune
Read the full abstract ▾
All authors R E Zera Marveen Lyngkhoi, Abhivansh Gupta, Vidushee Vats, Ram Mohan Rao Kadiyala, Vinay Kumar Sankarapu, Pratinav Seth
Machine Learning
PDE-PFN: Prior-Data Fitted Neural PDE Solver
Jaehyeon Park, Mingu Kang, Dongseok Lee et al. (+5 more)
Motivated by the success of large
language models (LLMs) with broad generalizability and robustness to noisy or unreliable pre-training data, we explore in-context prediction of PDE solution fields using approximate pre-training data. In addition, inspired by the posterior predictive mean-based inference mechanism of the in-context learning in prior-data fitted networks (PFNs), we propose PDE-PFN, a prior-data fitted neural solver that approximates the posterior predictive mean of PDE solutions via in-context learning. PDE-PFN builds on a PFN architecture with self- and cross-
attention mechanisms of
Transformer and is pre-trained on noisy approximate solutions generated by physics-informed neural networks, serving as diverse but not necessarily exact priors. Through experiments on a range of two-dimensional PDEs, we demonstrate that PDE-PFN achieves empirical generalization across heterogeneous equations, robustness under noisy priors, and zero-gradient-update in-context inference capability. PDE-PFN achieves competitive or improved predictive accuracy relative to the evaluated baselines while supporting flexible context and query representations.
Read the full abstract ▾
All authors Jaehyeon Park, Mingu Kang, Dongseok Lee, Woojin Cho, Kookjin Lee, Anthony Gruber, Youngjoon Hong, Noseong Park
Computer Vision
LETHE: Defending against Unauthorized Image-to-Video Generation via Value Diversity Collapse
Boheng Li, Yueming Xing, Yiming Li et al. (+2 more)
Image-to-
video (I2V)
diffusion models can animate a single reference image into high-quality, subject-consistent
videos, posing threats of misuse as anyone can turn personal photos into fabricated
videos without consent. A promising solution is to proactively add an
adversarial perturbation to the image before release, such that
videos generated from it would be unfaithful or visually corrupted. However, due to an insufficient understanding of what governs how subsequent frames read the reference image in modern I2V models, existing protections struggle to offer strong protection under common perturbation budgets. In this paper, we approach the problem by exploring this underlying reading mechanism. Through a convex
geometric analysis, we theoretically identify the spread of the reference value vectors as a key factor governing this reading process, and empirically find that even a mild contraction of this spread can severely degrade the generated
video, making it an efficient lever for protection. Guided by this insight, we propose LETHE, a novel protection method that directly disrupts this reading process. LETHE pairs two complementary objectives, where a value collapse objective drives the reference value vectors toward uniformity, while an
attention attraction objective draws subsequent frames toward these collapsed
tokens to amplify corruption. We further construct two human-centric I2V protection
benchmarks and conduct extensive experiments with multiple modern I2V models, which demonstrate the state-of-the-art protection performance and versatility of LETHE.
Read the full abstract ▾
All authors Boheng Li, Yueming Xing, Yiming Li, Jie Zhang, Tianwei Zhang
High Energy Physics (Phenomenology)
Chiral Symmetries in the Standard Model of Particle Physics and its Extensions
Sergio Ferrando Solera
Chiral gauge theories are a central focus of research in theoretical particle physics due to the parity
symmetry breaking observed in the Standard Model. Left-Right Symmetric Models (LRM) explain the breaking of parity through a Spontaneous
Symmetry Breaking Mechanism, predicting new $W_R$ and $Z_R$ gauge
bosons. The absence of direct BSM physics detection at the LHC has driven research towards precision observables and Effective Field Theories (EFTs), particularly the SMEFT framework. Precision radiative calculations in
chiral theories using dimensional regularization require a $D$-dimensional $\gamma_5$ prescription, such as the mathematically consistent Breitenlohner-Maison-'t Hooft-Veltman (BMHV) scheme. However, this scheme introduces gauge
symmetry-breaking terms that obstruct Slavnov-Taylor identities. A primary objective of this thesis is to provide a pedagogical introduction to cancelling these spurious anomalies via algebraic renormalization, and to derive the one-loop
symmetry-restoring finite counterterms required to renormalize the four-
fermion operators of the $\nu$SMEFT. Furthermore, the thesis explores UV completions like LRMs embedded in other BSM theories like the Grand Unified Theories. This
embedding implies that no discrete
symmetries remain at the TeV scale and allows LRMs to be treated as EFTs with the minimal scalar content needed to perform the Spontaneous
Symmetry Breaking Mechanism. The second main objective is to derive robust lower bounds on the masses of the new gauge
bosons and the LR scale using
collider observables. This analysis is performed for both the effective and traditional LRMs, ensuring that the results are independent of other parameters of the theory.
Read the full abstract ▾
Algebraic Geometry
Gamma conjecture II and explicit Stokes data for a quadric projective bundle
Xiaobin Li
We study the connection problem for the
quantum differential equation of the Fano fourfold $X=\PP_{Q^3}(\OO\oplus\OO(1))$. The problem is to identify its canonical asymptotic solutions with exceptional objects and compute the associated Stokes data. We use exact Mellin–Bessel integral representations and uniform vector asymptotics on sectors wider than $\pi$, followed by analytic continuation that preserves the Gamma framing. Cohomological
chiral symmetry and mutations of exceptional collections put all eight sections at a common phase. This proves Gamma
conjecture II for $X$ and determines a full exceptional collection and its Stokes and central connection
matrices at the unit Novikov point and a specified admissible phase. The
chiral symmetry also gives a flat symmetric bilinear pairing, an orthogonality identity for the Stokes
matrix, and an integral unipotent square root of the monodromy
matrix for a full turn. We obtain explicit formulas for the connection data at the opposite phase; on the Gamma
lattice, the half-turn acts by tensoring with a square root of the anticanonical bundle. The same spinor integral transform constructs two canonical Gamma sections for $\PP_{Q^{2m+1}}(\OO\oplus\OO(1))$, $m\geq1$, at positive parameters with distinct spectrum. Their sectors have opening greater than $\pi$, and their leading constants are independent of the dimension.
Read the full abstract ▾
Artificial Intelligence
Mengdi Liu, Chenghao Jia, Hong Chang et al. (+1 more)
Designing tool-integrated
reasoning agent systems for complex chemical tasks remains a fundamental challenge. Although recent chemical
agents have demonstrated promising performance through the integration of external tools, they are still built on the autoregressive paradigm, which is inherently constrained by causal
attention and left-to-right generation. As a result, they often struggle with bidirectional molecular understanding, global coordinated tool planning and multi-source evidence integration. To address this challenge, we introduce ChemDiffAgent, the first masked
diffusion language agent for tool-integrated chemical
reasoning. It reformulates chemical
reasoning as iterative denoising over interleaved
reasoning and tool-use trajectories, enabling tool-use decisions to be jointly determined under full-trajectory context and thereby better aligning the
agent's decision boundary with its knowledge boundary. We further analyze how this paradigm benefits agentic
reasoning through harder infilling subproblems during training and flexible arbitrary-order decoding during inference. Then we develop a two-stage post-training framework, comprising agentic supervised fine-tuning and variance-reduced preference optimization, to enhance tool-use and
reasoning capabilities. We also construct a comprehensive
benchmark covering six chemistry tasks across both single-turn and multi-turn tool-use scenarios. Under matched budgets and evaluation, ChemDiffAgent consistently outperforms autoregressive
agents, achieving an overall improvement of approximately 20%. Notably, our 8B model achieves performance comparable to, in some cases surpassing, significantly larger frontier models, such as GPT 5.5 and Claude Opus 4.7. Code is available at https://github.com/bilemond/ChemDiffAgent.
Read the full abstract ▾
All authors Mengdi Liu, Chenghao Jia, Hong Chang, Shiguang Shan
Astrophysics of Galaxies
The Silk Nebula: A Newly Discovered Giant [OIII] Emission Filament
Yi-Fei Wu, Songyu Jia, Chao-Jian Wu et al. (+14 more)
We report [O\,iii] narrow-band imaging of an extremely large-scale nebula, the Silk Nebula, discovered by the SWONS (Super Wide-Field [O\,iii] Narrow-band Survey), together with multi-fiber medium-resolution spectroscopic data covering the H$\alpha$, [N\,ii], [S\,ii], and [O\,iii] emission lines, acquired with the Large Sky Area Multi-Object Fiber Spectroscopic
Telescope (LAMOST) as part of the Medium-Resolution Spectroscopic Survey of Nebulae (MRS-N). Based on 2,684 high-quality spectra, we measured its physical parameters, including relative line intensities, radial velocities (RVs), and full widths at half-maximum (FWHMs). The [S\,ii]/H$\alpha$ ratio of the nebula significantly exceeds the shock-excitation threshold of 0.4. Its location in BPT diagnostic diagrams and spatial morphology of [O\,iii] both support the optical emission is dominated by shock-heated gas. The oxygen abundance is near
solar and spatially uniform, consistent with a shock propagating into the ambient interstellar medium. Because [O\,iii] is the only emission line that clearly traces the nebular morphology, we base our distance estimate exclusively on the [O\,iii] kinematics, obtaining a fiducial distance of $\sim$160 pc. At this distance, the nebula has a physical extent of $\sim$28 pc along the north-south direction. We suggest that the Silk Nebula is a large [O\,iii] filament possibly associated with a
supernova remnant, although a diffuse-ionized-gas (DIG) or ionized-bubble origin cannot be ruled out with the present data, and one of the largest angular-scale [O\,iii] emission structures documented to date. Future deep X-ray observations and shock-modeling efforts are required to firmly establish its nature.
Read the full abstract ▾
All authors Yi-Fei Wu, Songyu Jia, Chao-Jian Wu, Chao Zhang, Li-Yue Zhang, Shu Wang, Juan-Juan Ren, Quzhen Dongyang, Hong-Yun Qiu, Zhi-Hong He, Jichen Zhang, Guo-You Sun, Wei-Guo Zhang, Jian-Jun Chen, Hong-Liang Yan, Xiang-Song Fang, Hong Wu
Machine Learning
DuoRate: Two-Rate Per-Expert Precision Allocation for Mixture-of-Experts Serving
Mengqi Guo, Zhe Wei, Yuan Yuan et al. (+3 more)
Serving a Mixture-of-Experts (MoE) model pays for expert precision twice: the resident footprint pays for every expert the router can reach, while each batch-one decode step pays only for the experts it selects. These are two different rate functionals over the same per-expert widths, yet published per-expert MoE allocators constrain only the first. We call the difference the store-read gap; in the released BitsMoE checkpoint, 2.54 bits per weight (bpw) of storage become 7.50 bpw of decode traffic. We propose DuoRate, which stores once and reads per workload. Offline, it stores one uniform tier of a prefix-readable weight stream and bakes a distortion tensor. Online, it profiles a workload's own prefill without labels and solves a Lagrangian rate-distortion allocation under the decode functional; routing frequency provably cancels from every per-expert decision, and the solve costs one dot product per expert and a bisection. A deployment therefore keeps a single artifact that prefill reads at the stored rate and decode reads at the workload's own rate. On Qwen3-30B-A3B, DuoRate reading 2.3 bpw exceeds every 2-bit-class baseline by 6.3 to 47.0 points of average accuracy, and at 2.9 bpw it matches 4-bit formats that read 1.4-1.6 bpw more. The advantage is largest under long decoding: on GSM8K, DuoRate at 2.9 bpw matches or exceeds every 4-bit format and trails BF16 by only 0.25 points, and at 2.6 bpw it still matches or exceeds the 4.5-bit NF4 and FP4. Against controls in its own format, the allocation alone is worth 0.07-0.11 bpw of read rate. Code will be released upon publication.
Read the full abstract ▾
All authors Mengqi Guo, Zhe Wei, Yuan Yuan, Jiunn Bin Lim, Boyi Pan, Michael Bi Mi
Optimization and Control
Optimal Control of Prox-Regular Sweeping Processes with Variable Time
Tan H. Cao, Boris S. Mordukhovich, Trang Nguyen et al. (+1 more)
The paper is devoted to a broad class of optimal control problems governed by constrained sweeping processes with controlled dynamics, controlled prox-regular moving sets, and variable time included in the optimization process. Employing an advanced version of the method of discrete approximations, we verify its well-posedness with appropriate strong convergence of feasible and optimal solutions. This approach leads us to considering a parametric family of discrete-time controlled sweeping processes, which are reduced in turn to nondynamic mathematical programs with special types of
geometric and functional constraints. Invoking powerful tools of first-order and second-order variational analysis and generalized differentiation, we derive necessary optimality conditions for discrete-time and then—by passing to the limit over the diminishing step of discretization—continuous-time controlled sweeping processes. A crucial role in our device is played by explicit coderivative evaluations of set-valued mapping associated with the sweeping dynamics. In this way, we establish novel necessary optimality conditions for local minimizers of continuous-time controlled sweeping processes that contain absolutely continuous and BV adjoint arcs, local and global maximization conditions, $L^2$ active set multipliers, vector measures generated by sweeping and control constraints, and free-time transversality condition expressed through the averaged
Hamiltonian. For suitably selected multipliers, we also obtain nondegenerate forms of these conditions when the optimal trajectory starts on the boundary of the moving set. The efficiency of the obtained results is illustrated by some examples related to practical modeling.
Read the full abstract ▾
All authors Tan H. Cao, Boris S. Mordukhovich, Trang Nguyen, Thi Phung
Classical Analysis and ODEs
Sharp power-weight estimates for the fractional spherical maximal operator
Riju Basak, Surjeet Singh Choudhary, Chun-Yen Shen et al. (+1 more)
For $0\leq\alpha<n$, let $A_{*}^{\alpha}$ denote the fractional spherical maximal operator. In this paper we characterise the range of $\beta$ for which the $L^p-L^q$ power-weight estimate \[\left\|A_{*}^{\alpha} f\right\|_{L^q(|x|^{q\beta})} \leq C \left\|f\right\|_{L^p(|x|^{p\beta})}\] holds. In the diagonal case $p=q$, our work gives an affirmative answer to a
conjecture of Lee. We prove analogous results for the lacunary fractional spherical maximal operator.
Read the full abstract ▾
All authors Riju Basak, Surjeet Singh Choudhary, Chun-Yen Shen, Daniel Spector
Distributed Computing
Taking Tokenization off the Host for Agentic LLM Serving
Mohammad Siavashi, Pascal Schwinte, Ali Banaei et al. (+3 more)
Agentic workloads resend long contexts that barely change between calls. Prefix caching now skips most of their prefill, yet every request must still be tokenized in full before the cache can even be consulted. Tokenization, a CPU stage long treated as negligible, becomes the largest part of time-to-first-
token (TTFT), and a frontend that tokenizes for a whole GPU fleet limits how many GPUs it can feed. We present GraphTok, a GPU tokenizer that runs the entire encoding pipeline, from pre-tokenization through byte-pair encoding (BPE) merging to
token-ID assembly, as one replayable CUDA graph. GraphTok reproduces the reference
token IDs bit-for-bit on GPT-2, Llama 3, Qwen 2.5, DeepSeek-V3, and Gemma 3, hands short inputs to a CPU engine where the GPU cannot win, and plugs into unmodified vLLM and NVIDIA Dynamo. GraphTok encodes one million GPT-2
tokens in 1.7-3.2 ms including data transfers, and on Llama 3 it is 4.7-6.0x faster than gigatoken, the strongest CPU tokenizer. On TokTier's own single-request
benchmark, it is 2.6-3.9x faster than gigatoken and up to 18.1x faster than TokTier. In vLLM with a cached prefix, it lowers mean TTFT by 2.1-2.4x. The largest gain is capacity. Within a 2 s P99 target, one Dynamo frontend with GraphTok sustains 518 requests/s of 121k-
token traffic, the busiest-hour load of 56 GPUs, against 296 with gigatoken and 18.5 with the default tokenizer. Across 121k to one million
tokens, this is 1.75-2.5x the load gigatoken sustains, with a P99 up to 64x lower where gigatoken's tail breaks.
Read the full abstract ▾
All authors Mohammad Siavashi, Pascal Schwinte, Ali Banaei, Gerald Q. Maguire Jr., Marco Chiesa, Dejan Kostić
Signal Processing
Towards Signal Processing In Networks: Multiplierless Block Transforms in Data Plane
Shota Ono, Marika Morikawa, Kazuma Kaneko et al. (+4 more)
In this paper, we propose a signal processing concept, signal processing in networks (
SPIN), that executes signal processing tasks directly on programmable switches. Traditionally, signal processing tasks, such as denoising, decomposition, and dimensionality reduction, are performed on edge or cloud servers, i.e., raw streams are transmitted from sensors. This imposes a heavy processing load on edge/cloud servers and requires substantial bandwidth.
SPIN addresses this problem by performing signal processing on the forwarding path. As an initial realization of
SPIN, we implement a DCT-based block transform on a programmable switch. We redesign its computation using a lifting-based
lattice structure, multiplierless approximation, and pipeline partitioning across stages in a programmable switch. We verify at the compiler level that the proposed implementation satisfies the strict hardware constraints of the Intel Tofino2–specifically regarding pipeline stages, packet header vector, and ternary content-addressable
memory usage–and therefore enables line-rate processing. We further quantify the per-packet latency of the ingress match-action pipeline. The
SPIN-oriented preprocessing adds 46 cycles (approximately 31 ns at 1.5 GHz). The total ingress match-action unit latency is approximately 87 ns, which remains consistent with the low-latency operation targeted by
SPIN. Furthermore, we implement the proposed block transform on a BMv2 switch using P4 and evaluate it as preprocessing for in-network machine learning. Experimental results demonstrate that the proposed method improves classification performance over unfiltered baselines.
Read the full abstract ▾
All authors Shota Ono, Marika Morikawa, Kazuma Kaneko, Junya Hara, Hiroshi Higashi, Shigenori Tani, Yuichi Tanaka
High Energy Physics (Theory)
General relativity and its modifications as effective field theories of $GL(4)$ breaking
Tommaso Viacava and Omar Zanusso
We discuss General Relativity and some of its modifications as effective field theories of the breaking of the affine general linear group using the coset approach and the Maurer-Cartan form. While the ultimate goal is to break the
symmetry down to the Lorentz subgroup, we argue that there are three interesting breaking cascades in which the affine general linear group is first broken to a larger subgroup that contains the Lorentz group. In all cases, the breaking of translations always generates Goldstone modes that can be interpreted as metric degrees of freedom. We argue that the effective field theories that emerge as a result of the different breaking patterns are all dynamically equivalent at the lowest order of the derivative expansion, e.g., using Einstein-
Hilbert-like actions, but can differ beyond the leading order.
Read the full abstract ▾