Networking
Papers
Fresh from arXiv, typeset properly. Highlighted words trade on the market.
All of arXiv
Monday 9 November
In managed multi-tenant clouds, RDMA's lack of control-plane (CP) resource isolation lets a noisy tenant exhaust shared NIC microarchitectural caches, degrading victim throughput by up to 96.4%. Data-path schedulers leave CP resources unprotected, and hardware approaches require FPGAs or SmartNICs rather than the commodity RNICs operators already deploy. We present MTRDMA, a lightweight user-space CP isolation mechanism for commodity RNICs, applicable to managed runtimes such as containers and sandboxed processes. MTRDMA intercepts library-level CP operations to enforce hierarchical quotas while preserving the zero-copy data path, using atomic shared-memory quota state with per-process cached admission (${\sim}$0.13\,$\mu$s), a hybrid garbage collector that reclaims orphaned quota within seconds, and rdma-cgroup caps with ledger–kernel reconciliation to bound and detect residual tampering. On ConnectX-6 RNICs, MTRDMA cuts victim degradation under Queue Pair (QP) cache exhaustion from 94.5% to 0.08%, extends protection to Memory Region (MR) cache exhaustion, and scales to 256 concurrent tenants with negligible data-plane overhead.
Atomic Physics
Rapid high-purity loading of barium ion via relayed selectivity from naturally abundant source
Trapped ions is one of the promising platforms for realizing quantum information processing and quantum simulation. Selective and rapid loading of a specific isotope lays the critical foundation for such purpose particularly when scaling up the system size. Here we demonstrate a new method for rapid isotope-selective loading of $^{137}\mathrm{Ba}^{+}$ or $^{135}\mathrm{Ba}^{+}$, two types of promising ion qubits, from naturally abundant source using a three-step, ultraviolet-free photoionization scheme. Neutral atoms are selected on the narrow 791 nm $6s^2\,{}^1S_0\rightarrow6s6p\,{}^3P_1^o$ transition, resonantly transferred at 554 nm to $6s6d\,{}^3D_1$, and ionized using an existing 532 nm Raman beam. A spectral-overlap analysis based on isotope and hyperfine shifts shows relayed selectivity that the second step can further suppress undesired-isotope excitation and no contaminant ion is detected during our entire experiment. Besides, the resonant drive of the second step excitation yields a prominent loading rate of the selected species. This strategy using multi-step excitations to establish relayed selectivity can be extended to high-purity loading of other species of ions.
Read the full abstract
Sound
MUTE: Multi-Level Alignment Uncoupling Against Talking-Head Exploitation for Voice Protection
Talking-head generation models, which synthesize realistic facial animations from audio, are increasingly vulnerable to misuse in multimodal deepfake scenarios. Protecting such systems remains challenging, as talking-head models fundamentally rely on precise audio-visual alignment, while effective audio protection methods remain largely underexplored. In this work, we propose MUTE, an audio protection framework that uncouples the underlying audio-visual alignment through a multi-level strategy. Specifically, MUTE combines (1) Alignment-level Disruption, which directly perturbs cross-modal attention to disrupt structural alignment, and (2) Representation-level Degradation, which perturbs temporal audio embeddings to degrade their representations. To improve robustness, perturbations are constrained in the STFT domain to high-energy regions, making them resistant to post-processing, denoising, and adaptive attacks. An optional speaker-level objective further mitigates potential bypass via TTS-based resynthesis. Extensive experiments demonstrate that MUTE consistently degrades lip synchronization across both white-box and black-box talking-head models while preserving perceptual audio quality. The proposed method remains effective under real-world transformations and can be combined with image-based approaches to provide stronger multimodal protection. The source code and project page are available at https://github.com/jindl465/MUTE.
Read the full abstract
Nuclear Theory
Role of Wigner SU(4) Symmetry in Low-Energy Spin-Isospin Transitions of Three-Valence-Nucleon Systems on LS Closed Cores
We investigate low-energy Fermi (F) and Gamow–Teller (GT) transition strengths in three-valence-nucleon systems from the viewpoint of Wigner SU(4) symmetry. For $^{19}{\rm F}$–$^{19}{\rm Ne}$, shell-model calculations show that the ground states contain a large component of the SU(4) representation whose spin-isospin wave function is fully antisymmetric. By varying the spin–orbit splitting in the shell-model Hamiltonian, we demonstrate clear correlations of the SU(4) overlap with both the GT strength to the ground state and its fraction of the total GT strength. For $^{7}{\rm Li}$–$^{7}{\rm Be}$, shell-model calculations show that the $3/2^-_\mathrm{g.s.}$ and $1/2^-_1$ states both belong predominantly to the same supermultiplet as the $^{19}{\rm F}$–$^{19}{\rm Ne}$ ground states. The observed strong GT transitions to these states are therefore interpreted as transitions within the same SU(4) multiplet, split by the spin-orbit interaction. Thus, Wigner SU(4) symmetry provides a unified interpretation of low-energy GT strength concentrations in three-valence-nucleon systems. In particular, the concentration of both F and GT strengths in the same $^{19}{\rm F}$–$^{19}{\rm Ne}$ ground-state transition is naturally understood as a three-valence-nucleon extension of the SU(4) mechanism underlying the low-energy strong GT peak in $^{18}{\rm O}$–$^{18}{\rm F}$.
Read the full abstract
Computation and Language
When Audio and Video Mislead: Evaluating Multimodal AI for Conversational Emotion Recognition
Multimodal Emotion Recognition in Conversation (MERC) combines lexical, acoustic, visual and conversational cues, but accepting audio and video does not establish their effective or robust use. We audit off-the-shelf Omni models while holding the target transcript and prior textual context fixed and adding, removing or replacing target audio and video. In the primary evaluation of Qwen2.5-Omni-7B, Baichuan-Omni-1.5 and MiniCPM-o 2.6 on MELD and the full five-session six-class IEMOCAP set, T+A improves Accuracy in all six model-dataset pairs, with four gains significant after false discovery rate (FDR) correction. Across three source mappings, unrelated A/V (RandA/V) lowers Accuracy relative to Full in all 18 comparisons; all 18 remain significant under dialogue-cluster inference and BH-FDR correction, both on common-valid pairs and when invalid/OOM outcomes are treated as incorrect. Comparisons with modality-omission baselines further distinguish replacement penalties from mismatch-specific harm. A Qwen/MELD case study finds significantly greater gold-vs-best-candidate margin recovery in Late than Early or Middle under decision-query audio knockout, relative to an edge-count-matched Video-Sham control, using Holm-corrected tests. These results distinguish genuine-modality benefit from sensitivity and vulnerability to misleading input.
Read the full abstract
Data Structures and Algorithms
Tight Bounds for Sublinear Metric Steiner Tree
In the metric Steiner tree problem we are given an $n$-point metric space $(V,w)$ and a set $T\subseteq V$ of terminals, and we seek the smallest weight $\mathrm{OPT}$ of a tree that spans $T$ and may use points of $V\setminus T$. We study sublinear algorithms in the distance-oracle model, where the algorithm knows $V$ and $T$ upfront and can query the distance function $w$. The objective is to estimate $\mathrm{OPT}$ while querying significantly fewer than $n^2$ distances. The minimum spanning tree on $T$ estimates $\mathrm{OPT}$ within a factor $2$, and its weight can be approximated with $\widetilde{O}(n)$ queries. The natural question that has been studied is therefore the query complexity of estimating $\mathrm{OPT}$ within a factor strictly below $2$. Chen, Khanna, and Tan [SODA 2023] gave such an algorithm using $\widetilde{O}(n^{13/7})$ queries, and a lower bound of $\widetilde\Omega(n^{6/5})$. Later Mahabadi, Roghani, Tarnawski, and Vakilian [ITCS 2025] improved the algorithm to $\widetilde{O}(n^{5/3})$, and Mahabadi, Tarnawski, and Vakilian [ITCS 2027] improved the lower bound to $\widetilde{\Omega}(n^{7/5})$. In this work we settle the above question. We give an algorithm that estimates $\mathrm{OPT}$ within a factor $2-\eta_0$, for an absolute constant $\eta_0>0$, with query (and time) complexity $\widetilde{O}(n^{3/2})$. We also show that the exponent $3/2$ is tight.
Read the full abstract
Machine Learning (Statistics)
Trajectory Density Ratios under Weak Support: A Survey across the Sciences
Many problems across the sciences compare two stochastic processes through the sequences they generate. Reinforcement learning asks what a new policy would return from the trajectories of an old one; molecular simulation asks how much more a ligand binds from samples of the unbound state; climate attribution asks how a heat wave shifted under warming; and language-model post-training asks what a deployed agent will do from logged interactions. In every case, one quantity would answer all these questions at once: the ratio of the two probability laws over their generated sequences, the trajectory density ratio. This ratio exists only under strong conditions. When one process reaches states the other cannot, the ratio does not exist; when their overlap is thin, its variance grows without control with the sequence length; and when they concentrate on different lower-dimensional sets, the ratio is undefined altogether. Weak support of this kind is the norm for sequence data, not a sampling accident, and its price is paid in identifiability rather than in noise alone. This survey organizes the field around a three-level diagnosis of weak support and a map of the nine retreats that methods take when the full ratio is out of reach, from direct estimation and reweighting through staged decomposition, stationarization, and partial-identification bounds to smoothing, transport couplings, operator spectra, and direct estimation of a single functional. Each retreat keeps part of what the ratio offers and pays for the rest, and each application domain has standardized the retreat its data force. The survey closes with open problems: testable diagnostics at every level, partial identification at the trajectory level, a rate theory of time-amplified error, and benchmarks with ground truth.
Read the full abstract
Functional Analysis
Forelli–Rudin Type Operators in Infinite Dimensions, I
We introduce Forelli–Rudin-type integral operators on the open unit ball of \(\ell^2\). We completely characterize the boundedness of these operators between weighted \(L^p\) and \(L^q\) spaces for all \(p,q\in[1,\infty]\), as well as their membership in every Schatten class of operators between the corresponding weighted \(L^2\) spaces. We also establish trace formulas for these operators that cannot be obtained from their finite-dimensional counterparts by a formal passage to the limit as the dimension tends to infinity. Our results are not merely infinite-dimensional extensions of those obtained by Cheng, Fang, Wang, and the second author (2017), Zhao and Zhou (2022), and Ding (2022). Rather, they reveal new phenomena that are unique to infinite dimensions and arise from the fundamental geometric and measure-theoretic differences between finite- and infinite-dimensional spaces. A key ingredient in our proofs is the establishment of new Forelli–Rudin-type asymptotic estimates in infinite dimensions. These estimates also provide new insights into the development of function theory on the infinite-dimensional unit ball.
Read the full abstract
High Energy Physics (Phenomenology)
Axion-Photon Conversion in Variable-Domain Magnetic Fields and the Highest-Energy Photons from GRB 221009A
We investigate photon-ALP conversion as a simultaneous explanation of the ultra-high-energy photons observed from GRB 221009A by LHAASO and of the $\sim300$ TeV photon reported by Carpet-3. We model the magnetic field of the host galaxy as a set of domains with randomly oriented field directions and variable domain lengths, and propagate the photon-ALP system through the host galaxy, the intergalactic medium and the Milky Way. Fitting the combined LHAASO and Carpet-3 spectra, we find that a broad distribution of domain lengths raises the photon survival probability at the highest energies and accommodates the Carpet-3 point with photon-ALP conversion alone. The best fit, $m_a=3.85\times10^{-7}$~eV and $g_{a\gamma}=3.07\times10^{-11}\,\mathrm{GeV}^{-1}$, lies below the CAST bound, and parts of the $1\sigma$ and $2\sigma$ regions are also compatible with it.
Read the full abstract
Operator Algebras
An invariant higher index pairing under Morita equivalence for proper étale groupoids
For suitable proper étale groupoids, we study a part of the higher index that is invariant under Morita equivalence. We capture this part by a pairing between cyclic homology classes associated with isotropy groups and cyclic cohomology classes of the groupoid convolution algebra. We show that the pushforward is compatible with the maps induced by essential equivalences. This gives a numerical invariant depending only on the Morita equivalence class of the groupoid. In the case of a finite group action, we relate the resulting pairing to the equivariant index and to the Atiyah-Bott fixed-point formula.
Read the full abstract
Robotics
Thermo-VLA: A Thermal-Aware Vision-Language-Action Model with Gated Residual Modulation for Robotic Manipulation
Recent Vision-Language-Action (VLA) models have begun incorporating thermal sensing, but existing approaches typically treat thermal images as ordinary visual inputs, providing limited support for temperature-aware relational reasoning in manipulation. To address this limitation, we propose Thermo-VLA with a ThermoRes Encoder, which decomposes thermal images into cold and hot attribute representations and injects them into the RGB token through language-conditioned residual modulation, enabling temperature-driven target selection and continuous manipulation. We further construct six real-world robotic tasks on a dual-arm Unitree G1 platform, spanning basic grounding to relational reasoning. Experiments show that Thermo-VLA achieves an average success rate of 0.86 across four thermal-dependent tasks, outperforming a pi_0.5 baseline that directly incorporates RGB-thermal observations by 18 percentage points. Under unseen-object and scene variations, Thermo-VLA maintains an average success rate of 0.78, achieving the best generalisation performance among the compared methods. Beyond thermal sensing, we transfer the same residual modulation mechanism to colour attributes and observe consistent improvements in both simulation and real-robot experiments, indicating its potential to generalise beyond thermal attributes. Project website: https://j-x-chen.github.io/Thermo-VLA-web/
Read the full abstract
Computer Vision
AnyBreak: Towards Jailbreaking Any T2I Model for Any Concept
Text-to-image (T2I) models have demonstrated remarkable capabilities in image generation and editing, yet this progress has also introduced significant security risks. Prior work has shown that adversarial prompt can induce the generation of Not-Safe-For-Work (NSFW) content, including pornographic, violent, and graphic imagery. However, such attacks are typically mitigated by prompt filters and often rely on white-box assumptions. In this paper, we propose AnyBreak, a novel jailbreak framework that constructs adversarial images via multi-scale progressive semantic alignment. During the generation process, these adversarial images steer the generation trajectory in latent space towards unsafe data manifolds, enabling cross-modal transfer of malicious concepts from the text modality to the image modality. AnyBreak operates using only clean textual prompts and adversarial images at inference time, and can jailbreak any T2I models towards any malicious concepts in a fully black-box setting. We further introduce AnyBreak-V, an enhanced variant that incorporates a compensation term to bypass post-hoc safety checkers. Extensive experiments demonstrate that AnyBreak achieves high attack success rates while preserving visual fidelity and semantic consistency with the input prompts under black-box and clean-text constraints. Our findings reveal that existing T2I defense leave critical cross-modal attack surfaces insufficiently addressed. The code is available at https://github.com/luqi-glory/AnyBreak.
Read the full abstract
Machine Learning
FireGate: Bounding Behaviour Change Before Training in Continual RL Fine-Tuning
Fine-tuning a deployed language model on one task after another can change behaviour that should stay fixed, such as earlier skills, general chat or refusals. Today, such forgetting is found only after training, on the benchmarks someone chose to run. This work bounds, before training, how often its outputs change on chosen prompt distributions. FireGate places each task's update, such as a new LoRA, behind a gate, which is a classifier on the model's frozen lower layers that decides for each prompt whether the update is used. Where no gate fires, the model computes exactly what the old one computes. As a result, on any prompt distribution, every bounded behaviour score changes by at most the gates' firing rate there over a whole task stream. Calibrating each gate on unlabeled prompts of what must not change fixes that rate before training, with a PAC guarantee, whatever weights training produces. Forgetting is thus bounded by the gate's false-alarm rate, and its price is the gate's power: no update, gated or not, that changes as few protected outputs can change new-task outputs more often than the best test of that level tells new-task prompts from protected ones. On reinforcement-learning streams of up to five tasks, gated training keeps the earlier tasks and changes 3.3% of held-out chat responses, whereas ungated training changes 91.8%. Across seven models, the gain kept behind the gate follows its firing rate on new-task prompts, measured before training. The certificate covers the gated model as deployed, on the prompt distributions sampled for calibration, not against an adversary who searches for prompts that fire the gate.
Read the full abstract
Robotics
CSG-Nav: Confidence-Aware Semantic-Geometric Fusion for Robust Exploration and Accurate Zero-Shot Object-Goal Navigation
Zero-shot object-goal navigation requires autonomous agents to explore unseen environments and locate target objects without task-specific training, posing challenges in perceptual reliability under observation uncertainty and physically grounded goal termination. Although foundation-model-based approaches improve generalization, they remain vulnerable to unreliable long-range perception, hallucination-induced directional bias, and semantically plausible but geometrically incorrect stopping decisions. To address these limitations, we propose CSG-Nav, a zero-shot navigation framework centered on confidence-aware multimodal information fusion. CSG-Nav integrates three complementary fusion mechanisms. First, semantic-source fusion integrates open-vocabulary detection results, detection confidence, scene observations, and relational reasoning to produce more reliable direction-level semantic scores and suppress hallucination-induced directional bias. Second, spatial-confidence fusion weights multi-view semantic evidence using a Gaussian decay function, suppressing unreliable distant observations while accumulating spatially consistent exploration preferences. Third, semantic-geometric decision fusion combines semantic target detection with depth-based 3D localization to support physically grounded stopping decisions. Experiments on HM3Dv1, HM3Dv2, and MP3D demonstrate competitive zero-shot performance, achieving success rates of 58.9%, 71.6%, and 44.3%. Fusion-level ablations and real-world deployments in indoor and outdoor environments further verify the effectiveness and practicality of CSG-Nav for robust navigation under uncertain perception. Code is available at https://xr-ui.github.io/CSGNav.github.io/
Read the full abstract
Machine Learning
MixRoute: Rethinking Single-Distribution Training for Generalizable Neural Routing
Neural Combinatorial Optimization (NCO) is a promising paradigm for solving routing problems, yet its generalization across diverse data distributions remains a critical bottleneck. Existing methods typically tackle cross-distribution generalization through complex multi-distribution training or intricate meta-learning schemes, implicitly presuming that single-distribution training yields a weak zero-shot baseline. In this paper, we demonstrate that the potential of single-distribution training has been substantially underestimated and can be unlocked by a simple architectural inductive bias: Mix Normalization. To approach this, we first construct a comprehensive benchmark spanning three representative routing problems and 179 fine-grained datasets, capturing diverse shifts in node coordinates, customer demands, capacity and time windows. Building on this benchmark, we propose MixRoute, a neural framework that adaptively combines normalization statistics at different granularities to mitigate varying distribution shifts. Trained exclusively on uniform distributions, MixRoute generalizes effectively to a wide range of unseen distributions without any distribution-specific adaptation. Extensive experiments demonstrate that MixRoute consistently achieves state-of-the-art zero-shot generalization. Our work revisits normalization as a parsimonious alternative for zero-shot generalization against complex training-adaptation schemes.
Read the full abstract
Machine Learning
ShiftCache: Fast and Exact Indel Effect Prediction by Reusing Shifted Reference Encodings
Sequence-to-function models such as AlphaGenome and Borzoi read up to a megabase of DNA and score a genetic variant by comparing their predictions for the reference and alternate sequences. Insertions and deletions (indels) are costly to score, because an indel usually shifts the unchanged DNA after it onto a different pooling grid, which prevents direct reuse of the reference's encoder activations. The AlphaGenome Atlas scored more than 100 million indels with three full model passes each, although the DNA after an indel matches the reference shifted by the indel's length. We present ShiftCache, an exact reuse system that encodes each shifted reference once and shares it among all variants scored in the same window with the same shift. ShiftCache copies every encoder output a variant can reuse and recomputes the rest in batched tiles. On held-out AlphaGenome windows shared by many indels, in strict fp32 on one A6000, ShiftCache is 3.5 to 6.5$\times$ faster than deduplicated dense execution and 1.6 to 2.7$\times$ faster than our adaptation of layer-by-layer reuse to indels. It is 3.5 to 4.1$\times$ faster than dense reimplementations of two published analyses with Borzoi and AlphaGenome, and 2.6$\times$ faster than dense execution with Enformer. Scoring each variant in its own centered 131 kb window makes ShiftCache only 1.36$\times$ faster than deduplicated dense execution, and highly diverse insertion lengths or many dispersed edits can make it slower than layer-by-layer reuse or dense execution. For encoders that meet our conditions, ShiftCache computes the same predictions as dense execution in exact arithmetic. On an A6000, every checked output of the default AlphaGenome and Borzoi runs matches dense execution within an absolute and relative tolerance of $10^{-5}$. With larger tile batches, with Enformer or on an H100, a small fraction of values exceeds this tolerance.
Read the full abstract
Artificial Intelligence
AutoQS: Automated Quantization Strategy Design with Large Language Models
Quantization reduces LLM deployment costs, but effective strategies must balance model quality and storage. LLM-driven search can automate strategy design through repeated proposal and evaluation. This process must identify useful strategies within a limited evaluation budget. The key challenge is that actual quantization is expensive, while inexpensive proxies may not reflect quantized-model performance. To address this challenge, we present AutoQS, an Automated Quantization Strategy Design with LLMs. AutoQS keeps weight precision fixed and searches group-size allocations across layers and operators under deployment constraints. Evaluation history guides coordinated LLM proposals, while deterministic validation ensures compatible allocations. An inexpensive quality proxy evaluates candidates, and an candidate repository retains alternative allocations to inform subsequent proposals. Periodic actual quantization complements these proxy evaluations with measured quality and checkpoint storage, providing feedback for further search within the evaluation budget. Experiments on Qwen2.5-7B-Instruct show that AutoQS improves seven-task average accuracy over uniform-group GPTQ and AWQ by 0.70 and 1.12 percentage points, respectively, at four-bit weight precision.
Read the full abstract
Computational Complexity
Polynomial Sample Complexity of Noninteractive Simulation over Finite Alphabets
We study noninteractive simulation from a finite correlated source. Alice and Bob observe independent copies of the source and generate outputs without communication. We give a quantitative construction for arbitrary finite output alphabets: for every fixed source, a joint law that is simulable with arbitrarily many samples can be approximated using a number of samples polynomial jointly in the output alphabet sizes and the inverse accuracy. We first study binary outputs. A one-sided Wright–Fisher comparison gives a source-dependent explicit exponent and a finite-head Gaussian-threshold representation. We then handle arbitrary finite output alphabets through averaged influence regularity, simplex-preserving Gaussian moment lifting, and a common Gaussian projection for all correlated-prefix states. The sample exponent is independent of the output alphabet sizes. For a concrete biased binary source we certify an explicit exponent below $19$.
Read the full abstract
Machine Learning (Statistics)
Beyond Prediction Residuals: Conformal Regret Calibration for Robust Contextual Linear Optimization
Predict-then-optimize systems convert contextual forecasts into decisions, yet conformal robust optimization typically calibrates prediction residuals rather than downstream loss. For linear optimization, this parameter-space geometry can be badly misaligned with decision risk: large objective perturbations may preserve the optimizer, while small boundary-crossing perturbations may cause substantial regret. We introduce Conformal Regret Sets, a post-hoc, model-agnostic framework that conformalizes an optimization-derived regret surrogate and uses its calibrated level set for minimax-regret refinement. We first show that directly conformalizing exact regret, although sufficient to certify the nominal predict-then-optimize decision, yields a degenerate robust problem: absent an exogenous bounded support, every alternative decision has unbounded worst-case regret. We therefore use SPO+, a pointwise upper bound on exact regret, as the conformal score, obtaining distribution-free finite-sample marginal regret guarantees under exchangeability. For contextual linear programs, the resulting set decomposes into polyhedral pieces aligned with the normal fan of the feasible polytope. A vertex is active if and only if its predicted suboptimality gap is at most half the calibrated threshold $\eta$, enabling local, output-sensitive generation of all relevant pieces. Moreover, if $\Gamma$ is the predicted gap to the closest competing vertex, the nominal and refined decisions admit the sharper worst-case-regret certificate $(\eta-2\Gamma)_+$, with an exact early exit when $\eta\leq 2\Gamma$. Experiments on contextual fractional knapsack validate the geometry and certificates: continuous refinement reduces the $0.95$-quantile and $\operatorname{CVaR}_{0.95}$ of regret by $9.1\%$ and $13.1\%$ relative to nominal PTO, while vertex-restricted refinement changes only $10.0\%$ of nominal decisions.
Read the full abstract
Number Theory
An Analytic Evaluation of Euler Sharp Sums
We give a direct analytic evaluation of the Euler sharp sums $\zeta^\sharp(2,\{1\}^{d-3},\overline{2})$ for all even integers $d\ge4$, expressing them in terms of ordinary multiple zeta values of depth at most two. The proof reduces these sharp sums to a harmonic-number series and evaluates it using a generating-function identity, thereby proving the even-depth case of a conjecture of Xu and Zhao.
Read the full abstract
Cryptography and Security
Harness Matters: Measuring and Dissecting Security Resilience in Web Agents
Web agents extend LLMs from text generation to real-world web interaction by relying on execution harnesses that interpret user goals, perceive webpages, plan actions, and ground model decisions into browser operations. While prior work has recognized that agent security depends on more than the backend model alone, how harness design quantitatively shapes web-agent security remains insufficiently measured. To systematically study web-agent harness security, we first characterize web-agent harnesses as a four-stage execution workflow covering goal interpretation, webpage perception, reasoning and planning, and action execution. Based on this abstraction, we build WebTrap Farm, a systematic and extensible evaluation framework covering three representative attack categories: malicious user prompts, malicious prompt injection, and deceptive web design. To avoid conflating security with task capability, we use paired usability calibration tasks to separate genuine defensive behavior from capability limitations. Our evaluation shows that existing web-agent harnesses remain vulnerable to typical attacks, with some unsafe behaviors occurring even without explicit attack content, revealing harness-intrinsic security failures (RQ1). Statistical analysis further shows that both harness and backend model significantly affect security outcomes, indicating that deployed-agent security is jointly shaped by the harness and backend model (RQ2). Finally, we localize where defensive behavior emerges across workflow stages and identify more secure design mechanisms, providing actionable guidance for building more robust web agents (RQ3). These findings highlight the web-agent harness as an important security boundary and motivate moving beyond model-centric defenses toward harness-level security design.
Read the full abstract
Combinatorics
Five-Entropic Matroids Are Quinary
A finite matroid is representable over a field if its elements can be assigned vectors over that field such that the rank of every subset equals the dimension of the span of its vectors. For an integer q >= 2, a matroid is q-entropic if there exists a random vector indexed by its ground set, with each coordinate taking values in a common q-element alphabet, such that the entropy of every marginal subvector equals the rank of the corresponding subset multiplied by log q. It remains open to determine for which values of q every q-entropic matroid is representable over a field. For q=2 and q=3, representability over F_q has long been known. More recently, it was established that every 4-entropic matroid is representable over F_4 and that the 6-entropic matroids are precisely the regular matroids, that is, those representable over every field. We settle the previously unresolved five-symbol case, proving that a finite matroid is 5-entropic if and only if it is representable over F_5. For the nontrivial implication, we use Matúš's equivalence between entropic and partition representations. We show that a degree-five partition representation induces, on each embedded U_{2,4}-minor, an F_5-linear representation after coordinatewise relabeling, with a rescaling class independent of the chosen linearization. Using the foundation presentation of Baker and Lorscheid, we verify the compatibility relations among these local classes and thereby obtain a global F_5-representation.
Read the full abstract
Human-Computer Interaction
Executable Human Study Programs with Silicon Participants
We introduce HumanStudyProgram, a framework that represents a documented human study as an executable program with configurable LLM-based silicon participants. Its computational abstraction separates study logic from participant implementation through standardized APIs for presenting information and collecting responses. We implement a researcher-facing authoring tool that reconstructs source-linked study descriptions from papers and supplementary materials, supports researcher review, and generates executable HumanStudyProgram artifacts. Researchers can then configure silicon participants, run the programs, and analyze results from aggregate findings to individual responses. We demonstrate the approach through study cases and evaluate the tool through a hands-on expert review with 12 social science researchers. All completed the workflow and identified potential uses in understanding published protocols, prototyping studies, and examining AI behavior. We aim to improve the reproducibility of human studies by reconstructing their documented procedures as programs that others can reuse and build upon.
Read the full abstract
Probability
Khintchine's limit theorem for monotone convolutions
We prove that every weak limit of monotone convolution powers is monotonically infinitely divisible. In contrast, we prove that every probability measure on $\mathbb{R}$ can occur as a weak limit of monotone convolutions of an infinitesimal triangular array. We also discuss similar limit theorems for multiplicative monotone convolution on the unit circle.
Read the full abstract
Machine Learning
Pairwise distortion distribution for compression and quantization
Choosing the latent dimension is a critical challenge in data compression, where conventional hyperparameter optimization is costly and ad hoc. Intrinsic dimension provides a good alternative, which connects information-theoretic notions such as Renyi information dimension (RID) to rate-distortion dimension and random codebook compression. In this work, we introduce the pairwise distortion distribution (PDD) and its asymptotic limit, pairwise dimension (PD), which generalize the correlation integral to general distortion measures. We show that PD bounds RID, and PDD provides a lower bound for random codebook quantization performance, which can guide a practical use for compression, quantization, and dimensionality reduction applications.
Read the full abstract
Computational Complexity
On Non-Deterministic Representation of Boolean Functions
We study non-deterministic representation of Boolean functions by polynomials, introduced by (de Wolf, 2000). A polynomial $p(x)$ is said to be a non-deterministic polynomial for a Boolean function $f$ if for all $a\in\{0,1\}^n$, $p(a)\neq 0$ iff $f(a)=1$. The non-deterministic degree of $f$, denoted $\mathsf{ndeg}(f)$, is the minimal degree of a non-deterministic polynomial for $f$. In this work, we show the following main results: - Almost all Boolean functions on $n$ variables have non-deterministic degree at most $n/2 + O(\sqrt{n\log n})$. This confirms a conjecture of (Iyer et al., 2023). Together with the lower bound of (Iyer et al., 2023) it establishes a tight bound on the rational degree of a random Boolean function up to lower order terms. - We disprove a conjecture of (Kothari et al., 2026) about effective hypercube nullstellensatz for more than two polynomials. On the positive side, we also identify a constraint under which an effective hypercube nullstellensatz holds for many polynomials.
Read the full abstract
Artificial Intelligence
ChainCheck: Near-Perfect Evidence-Sufficiency Scores Can Conceal Sensitivity to Construction Artifacts – Paired Conditional Contrasts that Expose an Evaluation Confound
Evidence-sufficiency benchmarks ask whether a model can tell a complete passage set from one whose reasoning chain is broken. When broken sets are made by editing text, near-perfect sufficiency scores–and even positive crossed-control scores–need not demonstrate multi-hop reasoning. We present ChainCheck, an evaluation-validity audit that pairs validator-certified counterfactual cells with artifact-only baselines. In an exploratory crossed control on 2,214 HotpotQA pairs, a language-model scorer that never reads the chain reaches a nominal AUC of 0.997, and a one-string match reaches the maximum of the crossed chain-sensitivity statistic CS4 (0.500), while the 11 evaluated systems score from 0.308 to 0.484; the two statistics are different metrics. We therefore read positive CS4 as operational bridge-consistency sensitivity, not as semantic multi-hop reasoning. Every system is also sensitive to one extra edit, and the edits are detectable according to model-based naturalness ratings. On an independent HotpotQA set of 2,276 pairs, patterns pre-specified on MuSiQue and 2WikiMultiHopQA did not replicate: 5 of 6 trained readouts were edit-dominated, and the reference judge's selectivity interval [-0.007, +0.042] includes zero. Judge scores also depended on batch layout. The MuSiQue and 2Wiki records are public; the HotpotQA records and the fine-tuned checkpoints are not, so external reproduction is partial.
Read the full abstract
Computer Vision
3R-Adapter: Retrieval, Rewiring, and Refinement for Efficient Adaptation of 3D Reconstruction Model
Feed-forward 3D reconstruction models have demonstrated strong generalization under large-scale pre-training, yet they struggle in long-tail scenarios such as low-light reconstruction, joint human-scene reconstruction, and test-time adaptation. We observe relational attention structure collapse, manifested as the failure of cross-view correspondences, as a primary cause of these limitations. While adaptation offers a practical way to bridge this gap, existing parameter-efficient fine-tuning (PEFT) approaches mainly perform channel-wise feature adjustment and lack mechanisms to explicitly adapt relational structures, leaving correspondence failures unresolved. In this work, we formulate adaptation of 3D reconstruction models as a problem of correspondence recovery under domain shift and propose 3R-Adapter, a PEFT paradigm for long-tail 3D reconstruction scenarios. Our method consists of three components: Deformable Retrieval for task-adaptive relation candidate retrieval, Relational Rewiring for reconstructing cross-view relational structures, and Consistency Refinement for injecting the rewired relations into stable predictions. Extensive experiments on low-light reconstruction, joint human-scene reconstruction, and test-time adaptation demonstrate that our approach significantly improves robustness over existing PEFT methods while maintaining efficient lightweight adaptation. Our code is released at https://github.com/PKU-OV3-LAB/3R-Adapter.
Read the full abstract
Differential Geometry
Curve theory in $S^2\times S^2$ with applications to ruled surfaces in $S^3$
The Riemannian product $S^2\times S^2$ of two unit spheres admits several important geometric interpretations. It can be identified both with the complex quadric $Q^2$ and with the space $\mathcal{L}(S^3)$ of oriented geodesics in the unit $3$-sphere $S^3$. In this paper, we first develop a unified framework for the geometry of curves in $S^2\times S^2$. We introduce projection-regular curves in $S^2\times S^2$ and their curvature data, and prove a congruence theorem for these curves. As an application of this framework, regarding $S^2\times S^2$ as the complex quadric $Q^2$, we classify Kähler magnetic geodesics associated with the product Kähler–Einstein structure. For closed generic projection-regular curves, we establish a sharp Fenchel-type inequality. We then identify $S^2\times S^2$ with the space $\mathcal{L}(S^3)$ of oriented geodesics in the unit $3$-sphere $S^3$ and apply our curve theory to ruled surfaces in $S^3$. In particular, we prove that every nonsingular closed generic projection-regular ruled surface attaining equality in the Fenchel-type inequality is an embedded torus.
Read the full abstract
Machine Learning
Shaping Useful Noise: Energy Distributions Predict Visual Pretraining Quality
Synthetic data can scale visual pretraining, but current selection criteria only partially explain which datasets train strong representations. We propose using the energy profile of a dataset–the distribution of log-probability levels it occupies under a normalised real-image density model–as an audit and curation principle for synthetic data. To operationalise this view, we introduce Energy-Level Distribution Distance (ELD), which compares real and synthetic energy distributions. Real images span sparse high-likelihood scenes and dense low-likelihood textures, so useful synthetic data should cover this energy range rather than collapse to likely but low-variation samples. Across diverse synthetic datasets, ELD exposes mismatches missed by FID and predicts pretraining performance beyond FID, LPIPS variation, and feature-space recall, including across CNN and ViT self-supervised settings. We then use ELD to resample and mix synthetic sources, matching the real-image energy histogram while preserving diversity and improving downstream transfer.