Machine Learning
Adaptive Model Inversion Attacks Generalize a Privacy-Robustness Tradeoff
Shailen Smith, Rasmus Torp and Adam Breuer
In this paper, we show that standard evaluations of high-resolution Model Inversion Attacks (MIAs) significantly underestimate training-data
privacy leakage. State-of-the-art
privacy defenses, standard training techniques such as MixUp and
Adversarial Training, and undefended models all leak training images at rates 1.16 to 6.59 times higher on FaceScrub under simple adaptive changes to the attack, with the largest increases among defenses reporting the strongest
privacy. We further show that measured leakage depends on the feature basis of the external classifier used to evaluate reconstructions: for the same reconstructed images, an adversarially trained Inception evaluator identifies the targeted identity at different rates than the standard Inception evaluator. Our results suggest that standard MIA evaluation can mistake optimization and measurement failures for
privacy. These underestimated leakage rates also concealed a broader relationship between
privacy and
adversarial robustness. Once we adapt the attack and vary the evaluator, reconstruction leakage closely tracks
adversarial robustness across recent defenses and standard training regimes, suggesting that robustness provides an attack-agnostic proxy for reconstruction vulnerability that applies far more broadly than previously theorized. This raises an open question: can a practical defense reduce training-data reconstruction without paying a corresponding cost in
adversarial robustness?
Read the full abstract ▾
Machine Learning
Exact-Solution Volume and Length Generalization in Transformers
Yijia Jessica Zhu and David Chiang
Research on
transformer expressivity shows whether a
transformer is capable of solving a given task, but gives little indication of whether the solution, if learned, is generalizable to longer input lengths. We study this question through normalized exact-solution volume (NESV): the fraction of a bounded parameter region that achieves an exact solution on every input of length $n$. For fixed-width, single-layer
transformers with $\log n$-scaled
attention, we establish asymptotic bounds on NESV for four tasks: FIRST ($\Theta(1)$), MAJORITY ($\Theta(1/(n\log n))$), INDEX ($\Theta(1/n^3)$), and PARITY ($0$). These results are consistent with previous empirical results: the faster the exact-solution volume decays with input length, the harder it is to length-generalize on that task. Looking deeper into INDEX, our volume analysis reveals two error sources that grow with $n$. Consequently, we study a
transformer model that would structurally eliminate one of the terms, theoretically improving the NESV bound to $\Theta(n^{-1})$, and empirically achieving 85% accuracy when tested at $10\times$ the training length, compared with the 60% accuracy of the original model. We conclude that volume analysis may be a useful approach to identify concrete sources of length sensitivity and thus provide insights into task-specific model refinements.
Read the full abstract ▾
Artificial Intelligence
EIO-Agents: The Missing Semantic Layer for AI Agent Evaluation
Fouad Bousetouane
AI
agents are entering production in increasingly consequential environments without a shared semantic standard for what their evaluations actually mean. Scores, traces, judge outputs, and multi juror findings are increasingly used to justify readiness and release decisions, yet they often do not specify what evidence supports a claim, what that evidence can establish, or how the claim leads to a decision. We introduce EIO-
Agents, an open specification for interoperable AI
agent evaluation built on two layers. The Evaluation Intelligence Ontology (EIO) provides the semantic layer through typed evidence, versioned behavioral predicates, evidence contracts, claims, witness rules, proof status, recurrence, and computable derivations for metrics, findings, controls, and PASS, REVIEW, or BLOCK decisions. The Portable Evaluation Record (PER) provides the system of record: a canonical, content addressed representation of one evaluation that preserves the evidence to decision chain and can be re derived, explained, and verified. Scores summarize, juries interpret, and traces record, but none of them define what the evidence means or what it can prove. EIO provides that missing semantic contract, while PER preserves the resulting evaluation as a portable and verifiable system of record. As AI
agents assume greater operational responsibility, evaluation must become more than a collection of scores and verdicts; it must become an accountable artifact whose meaning, evidence, limitations, and decisions can be independently checked.
Read the full abstract ▾
Systems and Control
Belief-Informed Hybrid Control with Almost-Sure Target-Set Convergence
Clinton Enwerem, Saleh Kemal, John S. Baras et al. (+1 more)
Controlling a hybrid system to a target set under parameter uncertainty can require informative actions that temporarily drive the system state away from the specified set. We propose a belief-informed dual-control
algorithm that combines belief-space receding-horizon selection with an expected-decrease constraint on a nonnegative target-set progress function. To accommodate exploratory deviations, a scalar controller state bounds the constraint's cumulative slack. Our
algorithm's two-step lookahead selection uses predicted observations to update the parameter belief before evaluating the subsequent action's admissibility. Under distance-comparison bounds, correct conditional prediction, and recursive feasibility, we prove almost-sure target-set convergence at decision times and bound both the sum of expected progress-function values and the expected neighborhood-entry time. Upper confidence bounds extend these convergence guarantees to bounded model samples with summable error probabilities. In planar regulation with an unknown control direction, we verify recursive feasibility: the proposed two-step selection reduces the Euclidean state norm below 0.01 within 11 decisions for either sign, whereas a myopic one-step selection loses admissibility immediately after zero input. In a simulated bimanual assembly task, our
algorithm's one-step implementation uses clearance feedback to complete the assembly under nominal friction, yielding a 96.4% lower infinity-norm relative-position error than the reference-tracking baseline at the method's completion time. At lower friction, its posterior conditioning successfully detects an empty admissible set, whereas a fixed-prior alternative admits an action that violates the conditional decrease constraint. Project page: https://clintonenwerem.com/belief-hybrid-control/.
Read the full abstract ▾
All authors Clinton Enwerem, Saleh Kemal, John S. Baras, Calin Belta
Combinatorics
Quantum walks and graph operations
Joy Cooper, Homer De Vera, Hermie Monterde et al. (+2 more)
Let $U_X(t)$ be the transition
matrix of a
quantum walk on a graph $X$ relative to its adjacency
matrix $A$ or the
Laplacian matrix $L$. This paper investigates the behavior of
quantum walks under Cartesian products, joins, and graph complements. We have two main goals. First, we characterize the conditions such that peak state transfer and pretty good state transfer are preserved under these operations, allowing us to construct new families of graphs admitting these properties. Our second goal is to analyze the relationship between the
quantum walks on a graph and its complement. We provide bounds for $f_{u,v}(t)=\big|U_{X^c}(t)_{u,v}-e^{it\delta}U_X(-t)_{u,v}\big|$ and $g_{u,v}(t)=\big||U_X(t)_{u,v}|-|U_{X^c}(t)_{u,v}|\big|$, where $\delta=-1$ when dealing with $A$ and $\delta=n$ otherwise. Note that $f_{u,v}(t)$ and $g_{u,v}(t)$ both measure the difference between the behavior of
quantum state transfer between vertices $u$ and $v$ in a graph and its complement. If $X$ is regular or $M=L$, then $f_{u,v}(t)$ is bounded above by $\frac{2}{|V(X)|}$. If $X$ is non-regular and $M=A$, then we utilize the main
eigenvalues of a graph to obtain an upper bound for $f_{u,v}(t)$ which depends only on $A$. We also use the bounding
matrix of the graph to give bounds for the Nordhaus-Gaddum type relations $|U_X(t)_{u,v}|+|U_{X^c}(t)_{u,v}|$ and $|U_X(t)_{u,v}|\cdot |U_{X^c}(t)_{u,v}|$. Finally, we demonstrate that most of our bounds are sharp for certain families of graphs.
Read the full abstract ▾
All authors Joy Cooper, Homer De Vera, Hermie Monterde, S.A. Talebpour, Xuzhu Wang
Computers and Society
From Algorithmic Marginalization to AI-Mediated Re-Centering: Can Culturally Grounded AI Bring Hakka Language and Culture Back into Mainstream Society?
Chen-Chi Chang
As large
language models increasingly mediate writing, translation, information
retrieval, and education, the technological support available to a
language may influence its position in contemporary social life. For minority and minoritized
languages, this raises a dual problem: inadequate functionality can encourage movement toward dominant
languages, while apparently fluent assistance can normalize culturally distinctive expression. This conceptual Research Note develops a framework connecting algorithmic cultural marginalization, cultural grounding, and AI-mediated cultural re-centering. Algorithmic cultural marginalization describes the interaction of representational asymmetry, functional exclusion, homogenizing transformation, and recursive feedback. Cultural grounding identifies interventions across resources, models, services, and governance. Cultural re-centering concerns changes in accessibility, actual
language use, public visibility, participation, and the circulation of new knowledge. Drawing on research on
language shift and
language technologies, the paper uses Hakka in Taiwan as an illustrative case and develops eight propositions for subsequent empirical investigation. Its central argument is that expanded AI access should be assessed alongside the retention of linguistic variation and community authority. The framework distinguishes technological availability from social use, and factual accuracy from cultural fidelity, without assuming that AI can substitute for intergenerational transmission. It provides a basis for investigating whether minority-
language AI expands meaningful domains of use or reproduces linguistic marginalization through convenience and standardization.
Read the full abstract ▾
Combinatorics
The Area Asymptotics of $(sn,n)$-Dyck Paths
Evan Conway and James Harbour
We study the total and average area of $(sn,n)$-Dyck paths:
lattice paths from $(0,0)$ to $(sn,n)$ that stay weakly below the line $x=sy$, counted by the Fuss-Catalan numbers. Generalizing a result of Merlini, Sprugnoli, and Verri for the case $s=1$, we derive an exact formula for the total area over all such paths. From this we obtain explicit upper and lower bounds for both the total and the average area, together with the corresponding asymptotics: for fixed $s$, the average area is asymptotic to $\sqrt{\pi s(s+1)/8} \cdot n^{3/2}$, while for fixed $n$ it is asymptotic to $sn \cdot Q(n)/2$ as $s$ grows large, where $Q(n)$ denotes Ramanujan's $Q$-function. Along the way, we confirm a
conjecture of Kotesovec on the asymptotics of a binomial sum that also arises in several other enumeration problems.
Read the full abstract ▾
Signal Processing
Radar Intelligent Detection with Coarse-Grained Labels
Chuanfei Zang, Yumiao Wang, Xingyu Chen et al. (+3 more)
Many deep learning based radar target detectors rely on range-
cell level labels for training, which are expensive to obtain. To reduce the labeling burden, this paper presents a training strategy that uses only range-window level labels. Specifically, two sub-echoes are randomly cropped from the same window echo, resulting in a known relative shift between them. A siamese
Transformer is then used to identify target-salient responses from one sub-echo and form pseudo labels, which are mapped to the paired sub-echo according to the known shift. In addition, a shift-equivariant consistency (SEC) constraint is imposed to align the predicted responses of the two sub-echoes after shift compensation. By exploiting shift priors within each labeled window, the proposed approach learns reliable detection without requiring
cell-level labels. Experimental results show that it consistently surpasses classical constant false alarm rate (CFAR) detectors and achieves performance close to range-
cell level labels training, while substantially reducing labeling cost.
Read the full abstract ▾
All authors Chuanfei Zang, Yumiao Wang, Xingyu Chen, Yanbin Wang, Guolong Cui, Xiaobo Yang
Human-Computer Interaction
Evaluating human-AI workflows for field research in viticulture
Niko Carvajal Janke, Daoyuan Jin, Shivranjani Baruah et al. (+4 more)
We assessed the value of two live human-AI interactions in a precision disease control project in California vineyards. The project tested whether 2021-2024 commercial scouting records and remote-sensing measurements across 140 hectares could support 2025 red-leaf symptom
forecasting for prioritized scouting and virus testing. In Workflow 1, Aleks v1, a multi-
agent research system, developed
forecasting models with iterative human refinement. We applied Aleks's 2024 vine-scale model to updated 2025 predictors and evaluated red-leaf forecasts against independent 2025 scouting. In retrospective simulations surveying 45% of all vine positions, adding model-informed row prioritization to adaptive scouting increased the encountered proportion of newly recorded red-leaf observations from 85.8% to 94.1%. Within-block scouting comparisons suggested the model mainly improved scouting allocation among blocks. Despite unreliable internal 2024 performance estimates from synthetic oversampling before train/test splitting, Aleks developed an informative vine-scale model in 145 minutes, increasing throughput and answering our research questions. In Workflow 2, we assessed whether higher model-score vines had more frequent virus detection, and whether Aleks could infer this sampling goal from a general
prompt with data and literature. Aleks's plan prioritized balanced vineyard and model score coverage, while our plan prioritized field efficiency and high-model-score oversampling. Aleks's and our plans yielded 41/50 (82%) and 97/100 (97%) sampled vines. Aleks's plan omitted instructions for replacing missing vines, limiting implementation and operational value. Five of 137 sampled vines tested positive for grapevine red blotch virus (model score ROC AUC 0.735). These findings support assessing AI interactions by how well they advance field research objectives under live, project-specific constraints.
Read the full abstract ▾
All authors Niko Carvajal Janke, Daoyuan Jin, Shivranjani Baruah, Nicholas Gunner, Jacob Maus, Yu Jiang, Kaitlin M. Gold
Hardware Architecture
CACHEFORGE: LLM-Guided End-to-End Generative Cache Replacement Policy for Performance and Hardware Efficiency
Kaushal Mhapsekar, Bita Aslrousta, Brijesh Kumar Bhayana et al. (+5 more)
Modern cache replacement designs saturate because they operate within fixed representational structures, hand-crafted and heuristic based feature-engineered predictors, or offline imitation models that cannot generate new decision logic on their own. At the same time, replacement is shaped by the causal interaction of prefetching, thrashing, spatial locality, and access-type behavior, producing an enormous design space that is difficult to traverse manually. Prior approaches typically rely on heuristics, parameter tuning, or imitation of an offline optimal policy, capturing correlations rather than synthesizing new mechanisms. As a result, their performance gains often plateau and they overfit under dynamic workload conditions. CACHEFORGE is the first framework to evolve cache-replacement policies end-to-end by
embedding a large
language model inside a governed hardware-aware loop. In each iteration, the LLM proposes new C++ replacement logic, the policy is evaluated under a trace-based CRC-2 ChampSim simulator, and the framework enforces feasibility through reward shaping, structural checks, dynamic mutation, temperature scheduling, and cross-policy crossover. This closed-loop generation-
evolution loop specifically designed for cache replacement policy enables the discovery of compact policies that satisfy hardware constraints while exploring algorithmic transformations beyond fixed predictor structures. Across SPEC CPU2006, CACHEFORGE outperforms all CRC-2 baselines. It improves the total hit rate by 27.36%, 19.69%, 13.72%, 13.15%, 11.83%, and 5.73% over MPPPB, ReD, Hawk-eye, SHiP++, LIME, and LRU, respectively. On
memory-intensive workloads, it increases IPC by 10.15%, 7.89%, 6.34%, 3.64%, 3.12%, and 2.71% over LRU, MPPPB, LIME, ReD, SHiP++, and Hawkeye.
Read the full abstract ▾
All authors Kaushal Mhapsekar, Bita Aslrousta, Brijesh Kumar Bhayana, Paula Contreras, Azam Ghanbari, Ethan Goodman, Anna Andriiko, Samira Mirbagher Ajorpaz
Medical Physics
A No-Go Framework for Coarse-Grained Effect Prediction in Boron Neutron Capture Therapy: Structural Identifiability Limits and Microscopic Coverage
Atsushi Fujimura, Shuichi Furuya and Yasuaki Ichikawa
We establish when coarse-grained observables uniquely determine an expectation-based biological response in boron
neutron capture therapy (BNCT). Microscopic target populations are represented by probability distributions, while retained observables are finitely many population-averaged linear descriptors. For an admissible class containing all finitely supported microscopic distributions, exact distribution-free identification holds if and only if the single-target response kernel lies in the span of the constant function and retained descriptors. Thus mean dose, mixed-field component means, and finite moment sets are not generally sufficient under unrestricted microscopic heterogeneity. We also derive irreducible ambiguity and approximation bounds, sharp partial-identification intervals for bounded non-increasing convex survival kernels, a treatment-ranking criterion, and lower-tail and high-exposure bounds. A fixed-total-exposure temporal model distinguishes transient re-sampling from persistent target-specific underexposure, and a marked-Poisson model shows that separate frequency and event-quality marginals need not determine survival. The framework clarifies which microscopic information is lost through BNCT coarse-graining while leaving room for approximation, partial identification, and prediction under independently justified model restrictions.
Read the full abstract ▾
Human-Computer Interaction
SENSE: State-aware Emotion Navigation Storytelling Engine
Yi Xia, Pablo Carrasco Velo, Mudit Paliwal et al. (+5 more)
This paper presents SENSE, a state-aware framework for generating playable branching visual novels with multi-track emotional
navigation. Integrating a state-based narrative architecture called MIND, a structure analyzer, and a path-aware context management module, SENSE produces narratives that are both structurally coherent and emotionally rich. From minimal high-level inputs, it generates multiple intersecting routes while preserving character consistency and narrative causality. Evaluations using LLM judges, affective metrics, and visual assessments indicate SENSE outperforms baselines in narrative diversity and robust asset integration, while preliminary human trials show directional improvements in emotional fidelity alongside comparable enjoyment.
Read the full abstract ▾
All authors Yi Xia, Pablo Carrasco Velo, Mudit Paliwal, Ibrahim Khan, Yifan Geng, Mustafa Can Gursesli, Juho Hamari, Ruck Thawonmas
Databases
Mechanizing Transactional Anomalous Patterns for Weak Isolation Levels
Long Gu, Si Liu and Hengfeng Wei
Recently proposed transactional anomalous patterns (TAPs) provide a semantic characterization of weak isolation levels and serve as the foundation for TAP-based isolation checking. Yet, these characterizations currently exist only as pen-and-paper proofs. We present the first machine-checked formalization of TAPs and their associated weak isolation levels in Rocq. We further establish machine-checked proofs of the corresponding TAP-based characterization
theorems. The mechanization process revealed two subtle issues in the original presentation: an underspecified transaction history model and an incomplete TAP characterization of the Read Atomicity isolation level that failed to capture violations of session guarantees. We address these issues by refining the definition of transaction histories and the corresponding TAP definitions, and re-establish all characterization
theorems in Rocq. These refinements have further been integrated into the weak isolation checker Plume.
Read the full abstract ▾
Fluid Dynamics
Vectorial discrete unified gas kinetic scheme for continuum compressible flows
Yinghao Ma, Ziyang Xin and Zhaoli Guo
A vectorial discrete unified gas kinetic scheme (V-DUGKS) is proposed for continuum compressible flows. In the vectorial kinetic framework, mass, momentum, and total energy are represented by coupled distribution functions with separate relaxation processes for momentum and energy transport, enabling an adjustable Prandtl number. All
equilibrium distributions are truncated at the second-order Hermite level. Since only second-order velocity moments are required to recover the compressible Navier-Stokes equations, fourth-order Gauss-Hermite quadrature is sufficient for exact moment evaluation. Eliminating third-order Hermite terms allows compact discrete velocity sets to be used uniformly for all distribution functions, reducing sensitivity to the numerical reference temperature and improving stability and efficiency over the scalar DUGKS. The scheme employs a finite-volume formulation with characteristic-based flux evaluation and trapezoidal collision integration. Numerical tests, including shock-tube, Shu-Osher, two-dimensional Riemann, and three-dimensional Taylor-Green
vortex problems, demonstrate its accuracy and robustness. For the three-dimensional Taylor-Green
vortex, V-DUGKS achieves a speed-up of about 2.2-2.9 under the same CFL constraint due to a larger allowable time step and simplified
equilibrium formulation. These results show that V-DUGKS provides a robust and efficient kinetic framework for continuum compressible flow simulations.
Read the full abstract ▾
Multiagent Systems
Joint Workflow and Prompt Optimization for User Behavior Simulation
Nipun B Nair Tongtong Wu, Hongzhi Yin, Hui Li et al. (+1 more)
User behavior simulation is the computational modeling of user interactions within information systems through the use of simulated
agents in place of live users. It supports system testing and evaluation, decision-making and
forecasting, and user experience design. Existing simulators rely on hand-crafted rules or domain expertise that transfers poorly across tasks. SWORD (Simulation-driven Workflow and
Prompt Optimization with Role-based Design) is introduced as a framework that jointly optimizes multi-
agent workflow
topology and natural-
language prompts. It is guided solely by a scalar task metric, without domain initialization or task-specific engineering. The experimental results demonstrate that SWORD achieves statistically significant gains over
prompt-only, workflow-only, and staged-optimization baselines under a controlled, identical-backbone comparison. Against the strongest published domain-specific baseline, SWORD further improves accuracy while using a smaller backbone model, substantially less training data, and a very reasonable API cost (\$4--\$6 for each
dataset). Beyond predictive performance, SWORD autonomously discovers domain-relevant signals, review-sentiment mapping rules and epidemiological decay priors, purely from scalar error feedback, establishing textual gradients as a mechanism for unsupervised feature-importance discovery in user behavior modeling.
Read the full abstract ▾
All authors Nipun B Nair Tongtong Wu, Hongzhi Yin, Hui Li, Weiqing Wang
Machine Learning
MS-ECG-FM: Towards a More Universal Electrocardiogram Foundation Model for Health Monitoring using Multi-source Contrastive Learning
Robert A. Lewis, I-Min Chiu, Kyle Verrier et al. (+7 more)
Electrocardiography (ECG) records the electrical activity of the heart, aiding diagnosis by detecting abnormalities in cardiac function. ECG foundation models have demonstrated promising results, but are limited by a reliance on ECG interpretation reports as their sole supervision. Because interpretation reports only capture the subset of waveform information routinely recognized by clinicians, this constrains representation learning to overlook the broader diagnostic signals present in ECG. We introduce a new ECG foundation model — MS-ECG-FM — that is trained through contrastive
alignment to multiple distinct clinical note types, including ECG, echocardiography, radiology, and discharge reports. We evaluate MS-ECG-FM on an extended set of ECG detection
benchmarks, showing that it comprehensively outperforms existing methods on the full span of conditions that ECG can detect, including in reduced-lead configurations. Different reports improve representations for different diagnostic domains, while multi-source
alignment captures their complementary information and produces consistently strong representations across clinically diverse tasks.
Read the full abstract ▾
All authors Robert A. Lewis, I-Min Chiu, Kyle Verrier, Karthik Jayaraman Raghuram, Francoise Marvel, Salar Abbaspourazad, Anshuman Mishra, Guillermo Sapiro, Andrew C. Miller, Joseph Futoma
Artificial Intelligence
Massive Activation Gating Channel in Large Language Models
Minjia Mao, Shi Chen, Bowen Yin et al. (+1 more)
Massive activations, a phenomenon in which a small number of hidden channels exhibit exceptionally large magnitudes, are pervasive in large
language models (LLMs). However, the mechanism by which a
token develops massive activations as it propagates through a pretrained LLM remains poorly understood. In this paper, we find that the emergence of massive activations is controlled by a single channel in the input
embedding to a spike feed-forward network (FFN). The position of this channel is fixed for a particular LLM. We name this channel the massive activation gating channel (MAGC). When the value of the MAGC is sufficiently large (or small, depending on the LLM), the output of the spike FFN exhibits massive activations. Examining six LLMs across four model families and different model sizes, we verify the existence and effect of MAGC. We further provide a theoretical explanation of the mechanism by which MAGC induces massive activations. When the value of MAGC is sufficiently large (or small), the output of a spike FFN asymptotically reduces to a quadratic form that mixes a few columns of the down-projection
matrix of the FFN. Since these columns exhibit the shape of massive activations, the output therefore exhibits massive activations.
Read the full abstract ▾
All authors Minjia Mao, Shi Chen, Bowen Yin, Xiao Fang
Human-Computer Interaction
When the Commons Appropriates a Large Language Model: How WikiVault Reshaped Korean Wikipedia
Inhwa Song, Sohyeon Hwang, Ted Yoo et al. (+1 more)
Large
Language Models (LLMs) have disrupted the balance between content production and quality assurance that sustains knowledge commons, leading many to prohibit or restrict their use. But what happens when a community instead appropriates an LLM-powered tool for its own needs? We investigate this question through WikiVault, an LLM-powered editing tool developed within the Korean Wikipedia community and used primarily for translation. Combining ten interviews, platform-scale analyses, and matched quasi-experimental comparisons, we examine how WikiVault reshaped knowledge production on Korean Wikipedia. We find the tool 1) drastically amplified the production capacity of a small group of experienced editors, producing longer and more widely viewed articles; 2) shifted work toward reviewing articles and importing content; 3) imported not only content but also editorial judgments from English Wikipedia. Our findings show how LLM adoption can rebalance the interdependent work that sustains knowledge commons, while illustrating how communities can learn from emerging technologies through use and adapt their governance accordingly.
Read the full abstract ▾
All authors Inhwa Song, Sohyeon Hwang, Ted Yoo, Manoel Horta Ribeiro
Computation and Language
DLoop: Looped Speculative Decoding
Geonmo Gu, Byeongho Heo, HeeJae Jun et al. (+4 more)
Speculative decoding accelerates autoregressive generation in large
language models. In each drafting stage, a lightweight draft model proposes
tokens that the target model subsequently verifies. With increasingly capable draft models, we find that the target model frequently accepts all
tokens produced in a drafting stage. A verification nevertheless follows each drafting stage, resulting in unnecessary target-model forward passes even when drafting could have continued. Adaptive draft length methods decide during decoding how many draft
tokens precede a verification, but they raise the speedup only for autoregressive draft models. For a parallel draft model, drafting further requires target-model hidden states for draft
tokens that have not been verified. We propose DLoop, a looped form of speculative decoding that adaptively performs multiple drafting stages before verification. DLoop continues drafting while the draft model remains confident and verifies all accumulated draft
tokens together. Loop-aware training keeps the draft model reliable in the additional drafting stages by exposing it to its own hidden states for unverified draft
tokens. By spending additional draft-model forward passes, DLoop reduces the number of target-model forward passes required for verification. Across diverse speculative decoding methods including EAGLE-3, DFlash, Domino, DSpark, and multi-
token prediction modules, DLoop improves the wall-clock speedup by 5 to 41 percent while preserving lossless decoding. Code will be available at https://github.com/naver-ai/DLoop.
Read the full abstract ▾
All authors Geonmo Gu, Byeongho Heo, HeeJae Jun, Yoohoon Kang, Sangmin Lee, Sangdoo Yun, Dongyoon Han
Computational Physics
A pressure-based coupled general synthetic iterative scheme for rarefied gas flow simulation
Yanbing Zhang, Yifan Wen and Lei WU
This work develops a pressure-based coupled (PBC) GSIS. Within each GSIS iteration, a conventional kinetic solver advances the velocity distribution function, from which non-
equilibrium constitutive relations are extracted. A PBC macroscopic solver then solves the steady macroscopic synthetic equations: continuity and momentum equations are solved as a coupled pressure-velocity system under fixed temperature, followed by a segregated temperature solve. A boundary treatment is constructed to maintain strict consistency between kinetic boundary fluxes and macroscopic boundary conditions throughout the macroscopic iterations. The proposed method is validated against direct simulation Monte Carlo (DSMC) data across multiple three-dimensional test cases, including tube-to-
vacuum flows, a dynamic gas lock configuration for EUV lithography, divertors in
nuclear fusion, and sphere flow at Mach numbers up to 10. Numerical results demonstrate that our GSIS-PBC preserves asymptotic-preserving properties and yields satisfactory accuracy across continuum-to-transitional flow regimes. For low-speed near-continuum internal flows, GSIS-PBC delivers wall clock time speed-ups of up to two orders of magnitude compared with the conventional iterative scheme, and substantially outperforms the original density-based GSIS for challenging low-Mach internal flow problems. For
nuclear-fusion divertor flows, GSIS-PBC is faster than the DSMC by four orders of magnitude.
Read the full abstract ▾
Artificial Intelligence
Where Rules End and Judges Begin: Measuring the Judgment Boundary in Multi-Agent Systems Security
Shaswata Mitra, Raj Patel, Subash Neupane et al. (+3 more)
LLM-based multi-
agent systems (MAS) engage tools, share
memory, and delegate tasks, often encountering
adversarial content. Current defenses for MAS are typically evaluated in isolation, focusing on one attack type at a time, which can lead to costly and hard-to-audit outcomes. This study organizes defenses into five principles, implementing them as DEFER1 (DEterministic-First Enforcement with Residual judgment), which includes a cascade of 28 checks that blocks what it can and refers the rest to a panel of four judges. In independent testing across four domains, attack success rates drop from about 30.0% to approximately 3.0%, with 78% of blocked attacks handled by deterministic checks. Only a quarter of proposals reach the judges in the security-operations domain, illustrating that the rules provide security for attacks violating clear policies, while judges manage those that only misrepresent intent. Both systems have weaknesses, such as a risk-score approval gate that inaccurately approves most attack proposals but few legitimate ones, highlighting the challenges in assessing threats accurately.
Read the full abstract ▾
All authors Shaswata Mitra, Raj Patel, Subash Neupane, Sudip Mittal, Md Rayhanur Rahman, Shahram Rahimi
Superconductivity
Luttinger Liquid Behavior in a Single-Layer Nickelate La1.4Sr0.6NiO4
Yinqi Hu, Yidian Li, Xian Du et al. (+19 more)
The discovery of high-temperature superconductivity in
bilayer and trilayer nickelates has spurred intense interest in the Ruddlesden-Popper nickelates; yet the fundamental properties of the NiO2 layer remain obscured by interlayer coupling. It is therefore imperative to investigate the electronic properties of their single-layer counterpart to isolate the intrinsic physics of the NiO layer. In this work, we present a systematic study of the single-layer nickelate La1.4Sr0.6NiO4 using high-resolution angle-resolved photoemission spectroscopy (ARPES) and theoretical calculations. We reveal strong electron correlation effects, manifested by high-energy kinks in band dispersions and a pronounced orbital-dependent band renormalization. Interestingly, we observe a quasi-one-dimensional electronic structure characterized by straight
Fermi surface sheets along the diagonal momentum directions. Such square
Fermi surface
topology facilitates non-
Fermi liquid behavior consistent with the Luttinger
liquid model, as evidenced by the power-law spectral function, robust temperature scaling, and the observation of
spin-charge separation. Our results therefore not only unveil an exotic Luttinger
liquid behavior emerging from the unexpected dimensional reduction in an intrinsically quasi-two-dimensional nickelate but also provide a new perspective for understanding the intriguing physics in multilayer nickelates.
Read the full abstract ▾
All authors Yinqi Hu, Yidian Li, Xian Du, Wenxuan Zhao, Kaiyi Zhai, Senyao Zhang, Jiawei Shao, Mingxin Mao, Houke Chen, Jieyi Liu, Donghui Lu, Makoto Hashimoto, Fangyuan Zhu, Zhengtai Liu, Dawei Shen, Yaobo Huang, Zhongkai Liu, D. Prabhakaran, Yanpeng Qi, Yilin Wang, Yulin Chen, Lexian Yang
Machine Learning (Statistics)
Uniform Discrete Diffusion Models are Minimax Optimal for Estimating Distributions with Small Effective Support Size
Dongsun Yoon and Saptarshi Chakraborty
Discrete
diffusion models have emerged as a practically successful framework for generative modeling on discrete product spaces, yet their statistical generalization properties remain poorly understood. Discrete real-world data such as text or biological sequences often concentrate on a small fraction of the astronomically large ambient space because of semantic or physical constraints, but existing bounds fail to capture this distributional structure and instead scale with the size of the ambient space, giving rise to almost vacuous error bounds. We address this gap for uniform discrete
diffusion, one of the two dominant discrete
diffusion paradigms alongside masking
diffusion, by deriving statistical guarantees governed by the effective support size $s_n(P_0)$, a sample-size-dependent measure of distributional complexity. Given $n$ independent and identically distributed (i.i.d.) samples from an unknown data distribution $P_0$ on $[K]^d$, we show that, with appropriate choices of network size and hyperparameters, the expected total variation (TV) loss scales as $O(\sqrt{s_n(P_0)/n})$, while the expected Kullback–Leibler (KL) divergence is bounded by $O(\frac{1}{n}s_n(P_0)\log(eK^d/s_n(P_0))\log n)$. Furthermore, we show that the TV rate is minimax optimal and that the KL rate is minimax optimal up to a factor of $\log n$. Together, these upper and lower bounds show that uniform discrete
diffusion successfully avoids the curse of dimensionality for distributions with small effective support size: the TV error rate depends on the ambient state-space size only through $s_n(P_0)$, while the corresponding KL rate incurs only an additional logarithmic dependence on the ambient state-space size.
Read the full abstract ▾
Machine Learning
Jian Luo, Kehan Qi, Qingqiao Hu et al. (+5 more)
On-policy
distillation (OPD) has attracted growing
attention as an effective way to transfer capabilities from teacher models to student models. Recent studies further explore OPD as a tool for improving large
language model safety with promising results. However, these approaches typically assume that the teacher and training data are trustworthy. In this paper, we uncover an overlooked threat to OPD for safety: a safety-aligned but backdoored teacher can propagate its hidden malicious behavior to an initially clean student. Under our threat model, a poisoning rate as low as 3% results in an attack success rate (ASR) of up to 70% on the distilled student. We further identify two training choices that can amplify this risk. First, increasing the number of training epochs can lead to high ASR even at low poisoning rates. With only 10 poisoned samples, ASR reaches 67% after 16 epochs. Second, the commonly used top-k KL can accelerate
backdoor transfer, causing trigger-conditioned harmful behavior to emerge earlier than sampled-
token KL in most settings. Alongside these findings, we explore a simple mitigation, Lazy Defense, which clips KL rewards to make student updates less aggressive, limiting aggressive updates and slowing
backdoor learning. Experiments show that Lazy Defense delays
backdoor transfer in low poisoning rate settings. Together, our findings reveal that OPD can propagate
backdoors, highlighting the need to address the safety risks of OPD.
Read the full abstract ▾
All authors Jian Luo, Kehan Qi, Qingqiao Hu, Meilong Xu, Jiacheng Qiu, Weimin Lyu, Jiawei Zhou, Chao Chen
Numerical Analysis
Low-rank SI-DSA and GMRES-DSA for Radiative Transfer Equation with Adaptive Accuracy Control
Chenxi Han, Lukas Einkemmer, Wei Guo et al. (+1 more)
Recently, efficient sweep-based low-rank iterative solvers have been developed to reduce the computational costs of steady-state solves and implicit time integration of the radiative transfer equation (RTE). In particular, angular sampling based on the discrete empirical interpolation method (DEIM) or on error-indicated randomized greedy selection enables efficient, non-intrusive reuse of existing transport sweep implementations. However, randomized-greedy low-rank source iteration with
diffusion synthetic acceleration (SI–DSA) enforces uniformly tight accuracy tolerances across all source iterations, thereby oversolving the early iterations, and it does not explicitly reuse angular bases from preceding iterations. More critically, like its full-rank counterpart, this approach may lose effectiveness in multiscale problems with strong material discontinuities. We address these limitations through two complementary developments. First, we introduce adaptive accuracy control into randomized-greedy low-rank SI–DSA, reducing intermediate ranks and computational cost while maintaining the accuracy required for convergence. A two-stage hybrid sampling strategy can provide further acceleration by switching to DEIM-based sampling near convergence, once the preceding angular basis becomes sufficiently representative. Second, we develop an inexact flexible GMRES–DSA solver for the integral scalar-flux reformulation, in which sweep-based low-rank approximations are confined to the construction of the right-hand side and the
matrix–vector products. The scalar-flux Krylov basis is not compressed, thereby preserving Arnoldi orthogonality in exact arithmetic and allowing established inexact-GMRES analysis to guide adaptive accuracy control of low-rank operations. We present numerical results on
benchmark problems that demonstrate the effectiveness of the proposed methods.
Read the full abstract ▾
All authors Chenxi Han, Lukas Einkemmer, Wei Guo, Zhichao Peng
Robotics
SMART: Zero-Shot Sim-to-Real Articulated Object Manipulation via Large-Scale Synthetic Pretraining
Jicong Ao, Shuhan Jiang, Yuling Zhong et al. (+7 more)
The ability to interact with articulated objects is essential for
embodied intelligent systems, but collecting large-scale real-world demonstrations for these interactions remains challenging due to the precise contact and constraint-following motions involved. Although simulation provides a promising alternative, existing synthetic data efforts cover limited articulated-object categories, while general-purpose synthesis pipelines lack explicit designs for part-level semantics and articulation constraints, hindering agentic task generation and scalable synthesis of high-quality articulated-manipulation demonstrations. To bridge this gap, we introduce SMART, a scalable system leveraging large-scale Synthesized Manipulation demonstrations for ARTiculated-object manipulation. At its core, we develop SMART-Sim, a simulation platform with articulation-aware design that enables effective task generation and efficient demonstration collection. Building on SMART-Sim, we apply agentic task generation and design a scalable distributed synthesis system, using them to synthesize SMART-Data, comprising over 1M demonstrations across 44 atomic task types, 5
robot setups, and 2,507 articulated objects. The
vision-
language-action (VLA) model pretrained on SMART-Data shows competitive performance on simulation
benchmarks and achieves zero-shot sim-to-real transfer and scalable performance in real-world articulated-object manipulation tasks. This highlights the potential of synthetic demonstrations in providing effective and scalable supervision for improving VLA model performance in contact-rich articulated-object manipulation.
Read the full abstract ▾
All authors Jicong Ao, Shuhan Jiang, Yuling Zhong, Yanwen Liu, Yuhan Gao, Jiangyuan Zhao, Yang Zhang, Shiqiang Zhu, Chenjia Bai, Xuelong Li
Quantum Physics
Excitation of subradiant states in atom chains by surface plasmon waves
Giorgio Vimercati and Nicola Piovella
Subradiant states of a finite subwavelength atomic chain in free space offer a powerful resource for
photon storage. Yet their preparation and control remain challenging experimentally. A standard technique is to drive the atoms by a
laser and observe the florescence light of the long-lived subradiant modes of excitation after the
laser is switched off. However, these modes are highly degenerate and form an almost continuous band in the subradiant
manifold. Here we propose a new method of exciting a given subradiant state by coupling atoms to a surface plasmon wave propagating along the interface with a metallic film. By controlling the chain/surface distance we may load, store and release
photons in the atoms, making them a promising tool for future
quantum memories.
Read the full abstract ▾
Robotics
Silicon Language: A Robot-Native Knowledge Exchange Framework for Heterogeneous Robots
Yi Liu, Xianglin Meng, Chang Chen et al. (+1 more)
Reusing a capability across heterogeneous
robots still requires substantial human adaptation and verification: transferring a skill often means re-engineering interfaces, retuning parameters, and re-validating safety. We introduce Silicon
Language, a
robot-native knowledge exchange framework that treats the
robot as the active subject of its own capability
evolution. In this framework, a
robot that wants a capability encodes its own experience into knowledge packets, publishes them, retrieves peer packets, translates them for its own sensors and actuators, and reviews them through independent local trial. A receiver-side usability evaluation procedure lets each
robot decide for itself whether an external packet is useful, and progressive blending with automatic rollback is designed to reduce the risk of negative transfer when adopting it. The system combines three infrastructure layers (edge
agent, Silicon Transfer Protocol (STP), and knowledge hub) with a capability stack inspired by the human scholarly system. We report a 30-day proof-of-concept deployment at an above-ground simulated-mine laboratory in Yulin, with following trials on an outdoor sand road and an indoor factory floor. Two heterogeneous
robots encoded and translated three capabilities across embodiments through operator-assisted file copies mediated by the Silicon
Language translation layer; source-side trials of the
dust-locked following behavior were recorded on Taurus. Project records indicate that per-capability adaptation time dropped from days to hours; we present these figures as descriptive deployment records rather than controlled measurements. The deployment provides initial evidence for cross-embodiment knowledge exchange; fleet-level autonomous
evolution and controlled with/without-packet comparisons remain future work.
Read the full abstract ▾
All authors Yi Liu, Xianglin Meng, Chang Chen, Jingjing Fan
Robotics
OntoPlan: An Ontology-Grounded Scene Representation and Agentic Framework for Scalable Robot Task Planning
Hyeongwoo Nam, Woongje Cho, Juwon Kim et al. (+1 more)
Large
language model (LLM)-based
robot task planning is promising for open-ended instruction following, but degrades on long-horizon tasks in large environments. When spatial information is conveyed to the LLM through text, the model can fail to capture spatial context, and
token cost grows with environment size. Generating action sequences directly with an LLM also makes it difficult to satisfy the current world state and action preconditions. We address this with an ontology-grounded scene representation that aligns objects, spaces, relations, and states in a shared symbolic vocabulary for spatial
reasoning and task planning, and with OntoPlan, an agentic framework that interprets instructions, selectively retrieves task-relevant information, formalizes goals and constraints, and produces executable plans. Across 150 general tasks spanning five indoor environments and three scene scales, OntoPlan achieves 0.89 average task success, compared with 0.27 for the strongest baseline, while using 18.1k total
tokens per task on average, about 5.6$\times$ fewer than the most efficient baseline. These advantages persist as scene scale increases, whereas prior methods degrade more sharply in success and remain far more costly in
tokens. OntoPlan also responds appropriately to ambiguous or infeasible instructions by asking follow-up questions or reporting insufficient information rather than committing to invalid plans. Code available at https://github.com/namhyeongwoo/OntoPlan.
Read the full abstract ▾
All authors Hyeongwoo Nam, Woongje Cho, Juwon Kim, Jongeun Choi
Representation Theory
Block decompositions in the p-adic Langlands correspondence
Daniel Le, Matthias Strauch and Zichuan Wang
Let $K/\Qp$ be a finite extension, $\bG$ a connected reductive group over $K$, and set $G = \bG(K)$, considered as a $p$-adic Lie group. We show first that the Bernstein center $\cC_G$ of the category of solid locally $\Qp$-analytic representations of $G$ is isomorphic to the center of the locally analytic distribution
algebra $D^\la(G)$. We then consider the Emerton-Gee stack $\frX_{n,K}$ of rank-$n$ $(\vphi,\Gamma)$-modules over the Robba ring for $K$ and determine its connected components. The latter are in canonical bijection with the primitive idempotents of the ring $\cC_{\GL_n(K)}$. Moreover, under the assumption that the ring of global functions on $\frX_{n,K}$ has no non-zero locally nilpotent elements, we show that this ring is isomorphic to a Fréchet completion of $\cC_{\GL_n(K)}$.
Read the full abstract ▾