Peer Revenue Now accepting positions

Papers

Fresh from arXiv, typeset properly. Highlighted words trade on the market.

All of arXiv

Saturday 7 November

Computer Vision

Mental Imagery That Matters: Curing Latent Collapse in Multimodal Reasoning

Mental simulation, the ability to reason by constructing and updating internal visual states, is central to flexible human cognition, but remains elusive in current multimodal large language models. Latent multimodal reasoning offers a promising route toward this capability by interleaving text with continuous visual latent blocks, allowing models to maintain internal visual workspaces without decoding them into pixels. In this paper, we ask whether these latent visual states actually matter. Through a direct perturbation analysis across representative latent-reasoning MLLMs, we find that they often do not: dropping, replacing, or shuffling generated visual latents has only minor effects on the model's own final-answer probability, while equivalent perturbations to the linguistic reasoning trace are substantially more damaging. We call this failure mode Mental Simulation Collapse (MSC): the model appears to perform interleaved visual reasoning, but routes most answer-relevant computation through text. We further diagnose that MSC follows naturally from existing training objectives: visual alignment can make generated latents resemble helper-image embeddings, but it does not require the decoder to rely on them; subsequent text-only supervision further leaves the semantic role of the latent pathway unidentifiable. To address this, we introduce AnchorSim (Anchored Mental Simulation), a contrastive distillation framework that trains models to make their generated visual latents consequential. Across multiple visual reasoning benchmarks, AnchorSim improves performance, while diagnostic re-evaluation shows that visual perturbations now meaningfully affect predictions. Our results expose a fundamental failure mode of latent multimodal reasoning and demonstrate that mental simulation must be anchored to model behavior in order to matter.
Read the full abstract

All authors Mi Luo, Rohan Dugad, Zihui Xue, Alex Dimakis, Kristen Grauman

Nuclear Theory

Dissipative properties of isoscalar collective excitations in heavy nuclei

The dissipative properties of isoscalar collective motion in heavy nuclei are studied within a kinetic model of collective excitations in finite Fermi systems with a free moving surface. The model includes both the one-body damping produced by the interaction of nucleons with the dynamic surface of nucleus and the Landau damping mechanism. In this paper, we study the damping and dissipative properties of the isoscalar dipole collective modes and the isoscalar vortex octupole mode, which are formed due to dynamic effects of the nuclear surface. To study the damping mechanism of the considered collective modes, a multipole strength function is used in various approximations. It is shown that the damping of the considered isoscalar collective modes is due to a one-body damping mechanism. An analytical expression for the dissipation coefficient at finite frequency for the collective variable describing multipole oscillations of nuclear surface is obtained. It was found that the dissipation coefficient in the energy region of the considered isoscalar collective modes does not exceed the wall formula friction coefficient derived in the one-body dissipation theory. This result confirms the understanding of friction in the wall formula as a macroscopic limit of the one-body nuclear dissipation.
Read the full abstract

Analysis of PDEs

Nonexistence of positive solutions stable outside a compact set for the Lane–Emden system in a half-space

We prove that, in every dimension and for all positive exponents, the Lane–Emden system in a half-space with homogeneous Dirichlet boundary conditions admits no positive classical solution stable outside a compact set.

Optics

Dynamic Molecular Transformation Mediated Crystallization Regulation and Interface Passivation in Tin Based Perovskite Solar Cells

Tin-based perovskites are promising candidates for lead-free photovoltaics but suffer from labile Sn$^{2+}$ coordination, rapid crystallization, and severe interfacial defects. Here, we develop a dynamic molecular transformation using 1-acetylpiperazine (PX) to regulate crystallization and interface chemistry simultaneously. During nucleation and growth, the carbonyl group coordinates with Sn$^{2+}$ to stabilize precursor complexes, moderate supersaturation, and enable controlled crystal growth. Upon annealing, PX undergoes in situ molecular transformation, suggested to form amidinium-functionalized species that preferentially occupy surface A-site vacancies, forming a passivation layer and suppressing nonradiative recombination. This coupled bulk-to-interface evolution yields compact, highly crystalline films with improved optoelectronic quality. Devices achieve a power conversion efficiency of 15.18% and retain 94.1% of initial efficiency after 500 h of continuous illumination. This work demonstrates a dynamic molecular engineering approach integrating crystallization control and interface passivation, providing a pathway toward efficient and stable lead-free perovskite photovoltaics.
Read the full abstract

All authors Ying Tang, Zuhong Zhang, Hongzhuo Wu, Zhihao Li, Zhenhuang Su, Yonggang Yang, Hairui Liu, Feng Yang, Zhe Li, Luca Sapienza, Jorge Pascual, Shengnan Zuo, Bo Hou, Antonio Abate, Meng Li, Yufang Liu

Astrophysics of Galaxies

Time-Delay Recovery in AGN Reverberation Mapping via Mutual Information

Active galactic nuclei exhibit complex, nonlinear multiwavelength variability that standard cross-correlation techniques, sensitive only to linear covariance, may fail to capture fully. We apply time-lagged Mutual Information (MI) lag estimation, via the Kraskov-Stoegbauer-Grassberger $k$-nearest-neighbour estimator ($k = 5$) with a 150-iteration Random Subset Selection bootstrap, to the light curves of four Seyfert galaxies (Mrk 509, NGC 5548, NGC 4151, and NGC 4593) monitored across eight energy bands by Swift. In the UV/optical regime the MI estimator successfully reproduces the inter-band disc reprocessing lags established by the interpolated cross-correlation function, with performance degrading toward the optical where low variability amplitude produces poorly defined MI peaks. In the X-ray bands, where the physical X-ray/UV coupling is known to be weak, the estimator fails to recover consistent lags across the sample, with the notable exception of NGC 4593. Contextualising these outcomes within the multifractal characterisation of our previous work, we discuss how MI performance relates to the multifractal spectral width $\Delta\alpha$ of each band, and we propose $\Delta\alpha$ as a candidate diagnostic for anticipating where MI-based reverberation mapping is likely to succeed or fail in multiband datasets of active galactic nuclei.
Read the full abstract

All authors R. A. A. Souza, E. Hatziminaoglou, D. O. Fontinele, L. B. T. Santos, M. V. D. Souza, T. Stosic

Artificial Intelligence

Beyond Query Similarity: Safety-Conditioned Cache Routing for LLM Response Reuse

Semantic response caching reuses a cached response for a semantically similar query to improve LLM inference efficiency. However, query similarity alone can lead to unsafe reuse when safety conditions change. Existing approaches improve cache safety through stricter similarity checks, context-aware validation, or response verification, yet they may still miss safety conflicts between the cached response and the new request or introduce unnecessary verification overhead. We argue that safe and efficient semantic caching requires a reuse policy that makes the cached response's safety conditions explicit and uses them to guide the reuse decision with a balance between safety and utility. In this paper, we propose SafeCache, a safety-aware framework that separates candidate retrieval from safety approval and formulates semantic response reuse as a safety-conditioned routing problem. SafeCache matches each cached-response condition against the new request and aggregates the per-condition mismatches into an overall risk score. A separate contradiction guard prevents explicit conflicts from being diluted by strong matches. It further estimates matching uncertainty and calibrates the routing policy under a required unsafe-reuse target and a verifier-call budget to choose among direct reuse, selective verification, or regeneration. Experiments show that SafeCache achieves the better safety-utility trade-off among the evaluated baselines, enabling safer response reuse at comparable cache utility.
Read the full abstract

All authors Shilong Pan, Zhiliang Tian, Weiguo Chen, Zihan Chen, DongSheng Li

Strongly Correlated Electrons

Momentum-Resolved Kondo Hybridization of Nonsymmorphic Dirac-Derived Bands in CeAgBi$_2$

How symmetry-constrained conduction states acquire heavy-quasiparticle character is a central question in correlated quantum materials. Combining angle-resolved photoemission spectroscopy, Ce $4d \rightarrow 4f$ resonant photoemission, and first-principles calculations, we investigate the nonsymmorphic Kondo lattice CeAgBi$_2$. We identify quasi-two-dimensional Bi-square-net-derived bands consistent with calculated Dirac crossings and nodal lines. Resonant photoemission resolves a nearly dispersionless Ce 4$f$ band near the Fermi level, establishing Kondo heavy-quasiparticle states. Its pronounced momentum-dependent spectral weight at intersections with the identified Dirac-derived branches provides direct spectroscopic evidence for their participation in $c$-$f$ hybridization. Complementary measurements on independently cleaved samples and an effective hybridization model connect these branches to the Kondo heavy-quasiparticle spectrum. The combined results support preservation of nonsymmorphic Dirac-band connectivity under Kondo renormalization, showing how symmetry-constrained conduction states become constituents of correlated heavy quasiparticles.
Read the full abstract

All authors Yang Luo, Hao Liu, Zhong-Tuo Fu, Qi-Yi Wu, Bo Chen, Zheng-Tai Liu, Mao Ye, Da-Wei Shen, Chen Zhang, Ying Zhou, Cun-Hui Lyu, Jin-Dong Bai, Zi-Jie Xu, Hai-Long Deng, Zhen-Xing Wu, Jun-Jie Wu, Sai-Chao Deng, Yu-Xia Duan, Jian-Qiao Meng

Computer Vision

S2HAND: Accuracy-Preservation Trade-Offs for Hand Specialization in Whole-Body Pose Models

Unified whole-body pose estimators jointly predict body, face, and hand landmarks, but often degrade on small, blurred, or occluded hands. Attaching a standalone hand specialist introduces extra detection overhead, coordinate alignment, and wrist-stitching artifacts. We introduce S2HAND, a hand-specialized Sapiens2-5B framework that preserves the native 308-keypoint output interface. Specifically, S2HAND-MT adapts a lightweight 0.87M-parameter pose branch to route 40 finger heatmaps while strictly preserving the shared Transformer, auxiliary task heads, and the other 268 body, foot, and face heatmaps. S2HAND-Pose further fine-tunes three Fisher-selected Transformer blocks (267.13M parameters), trading exact preservation for stronger hand recovery: it lifts macro AUC from 57.83% to 65.73% across HInt TEST, COCO WholeBody VAL, and FreiHAND EVAL, and surpasses HInt-adapted Hand5/HRNet and stitched Hand5 systems on all five Native Pipeline benchmarks. On COCO, S2HAND-Pose improves hand AP by 12.89 points and body-plus-foot AP by 0.19 points. Across a 512-image multi-task audit, it maintains 99.60% foreground segmentation agreement with the public checkpoint, with quantified normal and pointmap drift. Its 2D coordinates also lower 3D mesh error in frozen Pose2Mesh transfer, while our optional TailGuard re-cropping module cuts S2HAND-MT's severe failure rate on NewDays from 13.45% to 10.43%. S2HAND thus resolves the whole-body hand bottleneck with either exact non-hand preservation (MT) or empirically audited trade-offs (Pose).
Read the full abstract

Algebraic Geometry

Birational Geometry of K-Moduli Spaces of Genus Four Curves

We study the wall-crossings for K-moduli spaces of (3,3)-curves on $\mathbb{P}^1\times\mathbb{P}^1$. Together with the boundary polarized Calabi–Yau moduli space, these spaces interpolate between the GIT moduli space of (3,3)-curves and the Baily–Borel compactification of the moduli space of K3 surfaces arising as cyclic triple covers of $\mathbb{P}^1\times\mathbb{P}^1$ branched along such curves. As applications, we describe the corresponding birational models in the Hassett–Keel–Looijenga program and realize the K-moduli space of degree one del Pezzo surfaces through explicit wall-crossings.
Read the full abstract

All authors Kenneth Ascher, Kristin DeVleming, Hongyu Luo, Yuchen Liu, Junyan Zhao

Computation and Language

Track or Follow? A Jev-Compatible Multi-Agent Trading System When Web News Lags the Market

LLM agents increasingly trade on the Web news they read, yet much Web news about listed firms is published after, and about, the price move it reports, so a reader can agree with the market by following it instead of reading the event. We ask whether LLM judgments of Web news track the market's reaction to the reported event or follow it. To answer, we built a multi-agent trading system whose agents share only point-in-time stores and answer typed questions through the interface of Jev, a hosted typed decision model; code sets every position under a proved bound on how far text can move it. In a hash-committed test of a Jev-compatible reader built from a pinned LLM's option probabilities, item judgments track announcement-window abnormal returns (weekly Spearman 0.119), also after the LLM's release, and event types agree with the issuers' 8-K filings (AUC 0.833); Jev's judgments of news published after its own release track the reaction as well (descriptive). Exploratory analyses show where the association comes from: 72% of earnings news items appear after the issuer's filing, contributor-written items often report the price move, and from the first trade after publication the association falls to 0.023. At a weekly horizon, the confirmatory test detects no rank information over code-only features. Under attack, decision cards traced 88% of successful attacks to the planted item. We propose that market-graded tests of Web-text readers report the criterion from publication, not only from the event.
Read the full abstract

All authors Yebo Feng, Guangyan Gan, Liangrui Pan, Ruichao Liang, Cong Wu, Jiahua Xu, Yang Liu

Computer Vision

Frequency-Structure Complementary Modeling for Hyperspectral Image Fusion

Existing methods mainly rely on convolutional aggregation or attention-based spatial-spectral interaction. However, conventional feature-domain attention does not explicitly characterize variation patterns along spatial-token and learned latent-channel coordinates, while spatially shared local operators lack the flexibility to adapt neighborhood aggregation to heterogeneous structures. To address these limitations, we propose a Frequency-Structure Complementary Modeling Network (FSCM-Net), which couples frequency-aware dependency learning with structure-adaptive local refinement. Specifically, a Spatial-Spectral Frequency Calibration Attention (SSFCA) module constructs spatial-token affinities from Fourier-interacted query-key representations and builds latent-channel affinities after input-conditioned frequency filtering along learned channel coordinates. The latent-channel response is then calibrated by a learned shared gain and combined with the spatial-token response. Complementarily, a Structure-Coherent Dynamic Feature Fusion (SC-DFF) module combines regularized structure-tensor cues from MSI features with context-modulated HSI detail representations to generate channel- and position-dependent aggregation kernels. This design makes local HSI refinement responsive to smooth regions, edges, and textures while maintaining spectral fidelity. The two modules are progressively integrated within a multiscale encoder-decoder to coordinate representation-level interaction and structure-adaptive reconstruction. Experiments on two simulated benchmarks and one real remote-sensing dataset demonstrate favorable performance against representative fusion methods in spatial detail reconstruction and spectral preservation.
Read the full abstract

All authors Song Qiya, Tian Siyu, Li Shutao, Dian Renwei, Sun Long, Liu Jinyang

Artificial Intelligence

TurnFlow: Benchmarking User-State Reasoning in Multi-Turn LLM Agents via Scene-Guided User Simulation

Large Language Models (LLMs) are increasingly deployed through conversational interfaces, yet strong single-turn performance does not imply reliability in multi-turn settings. Multi-turn assistance requires a model to maintain an evolving representation of user requirements and to reason over it before complying with later requests. Failures can therefore be subtle: models may fail to recognize a changed constraint or hidden conflict, or recognize it without enforcing it in the final response. Existing benchmarks primarily test fully specified prompts, long-context recall, or controlled domain environments, leaving open-domain requirement reconciliation undermeasured. We introduce TurnFlow, an interactive benchmark for evaluating whether models preserve, revise, and act on user requirements across turns. TurnFlow combines adaptive user simulation with verifiable judging, allowing conversations to unfold naturally while keeping hidden requirements and pass/fail criteria controlled. In our evaluation, nine LLMs achieve overall pass rates from 42.86% to 59.59% on 490 scenarios. Performance varies substantially across challenge types: models perform well on Stateful Multi-Constraint Following but remain considerably less reliable on Scope Creep and Goal Shift. These results reveal persistent limitations in reasoning over evolving user requirements across turns.
Read the full abstract

All authors Zhixun Chen, Ming Li, Anjie Liu, Xiaoyuan Zhang, Quanyu Dai, Huangxun Chen

Artificial Intelligence

Tail-risk-aware Agent Safety: Constraining Tail Risks Across Multi-step LLM Agent Workflows

LLM-based agents extend LLMs with tool calls, memory, planning, and retrieval to act in external environments, but they are vulnerable to attacks at different workflow steps. Existing defenses, including prompt engineering and post-training alignment, primarily improve LLM-only safety and often overlook risks that arise from multi-step execution. External verification adds guardrails to check each step separately, but it typically ignores how risks can propagate across steps. We argue that we should consider risk transitions across dependent steps through the agent workflow, rather than separate verification on a single step. In this paper, we propose Tail-risk-aware Agent Safety (TAS), a tail-risk-aware safety framework that models agent workflow safety as a sequence of dependent safety states and constrains the rare but extreme risks (i.e., tail risk) with a shared budget. TAS first models safety-state transitions sequentially along the workflow as an MDP to trace risk propagation, then measures a tail-sensitive risk to target rare but high-cost risks (i.e., tail risk), and constrains the risk with a shared budget to derive a step-wise safety strategy. By constraining tail risk rather than the overall average, TAS suppresses rare high-cost failures while preserving utility in normal cases. Experiments show that TAS achieves better performance compared with eleven baselines across three benchmarks.
Read the full abstract

All authors Shilong Pan, Zhiliang Tian, Zihan Chen, Wentao Zou, Weiguo Chen, DongSheng Li

Robotics

SimDuoReal: Object-Centric Bimanual Policy for Emergent Dexterity

Bimanual dexterous manipulation requires coordinated arm and hand control, yet specifying such coordination manually is challenging and often task-specific. In this work, we explore whether complex bimanual dexterity can emerge from a simple object-centric learning objective. We train reinforcement learning policies under a shared object-centric reward, varying their training environments rather than tuning separate rewards for each behavior. We find that the policies discover pre-grasp manipulation, grasping and regrasping, in-hand adjustment and handover to reach object-pose goals without hand-designed motion sequences. We then refine the experts on transition states and use DAgger and reinforcement learning to distill them into a single policy that composes these behaviors for diverse bimanual tasks. On real hardware, the same policy succeeds in 61 of 80 trials (76.3%) across four tasks with unseen objects and no task-specific fine-tuning. In simulated barcode scanning, our policy achieves 85.0% success, compared with 60.4% for two single-arm policies without handover. These results suggest that this objective provides a simple and scalable interface for learning general bimanual dexterity, shifting the burden from manually designing robot coordination to specifying what should happen to the object.
Read the full abstract

All authors Qiyang Yan, Yulin Liu, Akash Sharma, Haozhi Qi, Guanya Shi, C. Karen Liu

Dynamical Systems

Directional Ergodic Theorems, Ergodicity, and Weak mixing for $\mathbb Z^2$-actions

In this paper, we study Birkhoff and von Neumann type ergodic theorems for $\mathbb Z^2$-measure preserving systems along irrational directions. This leads us to a further investigation of directional ergodicity and weak mixing. Additionally, we show that our framework (dynamical definitions) is equivalent to the spectral definitions of directional ergodicity and weak mixing introduced by Robinson, Rosenblatt and Şahin \cite{RRS2023} for $\mathbb Z^2$-actions. We note that in their recent paper \cite{RS2026}, Robinson and Şahin established the equivalence between dynamical and spectral directional weak mixing for $\mathbb Z^2$-actions using an approach fundamentally different from the one employed in the present paper.
Read the full abstract

Machine Learning

Sharp Logarithmic-Product Bounds for List Replicability of Sign Matrices

List replicability measures how many accurate hypotheses suffice to contain the output of a learning rule with arbitrarily high confidence. We prove that every nonempty class of $m$ binary concepts on $n$ domain points has list replicability at most $805\ln(2m)\ln(2n)$. This resolves the logarithmic-product conjecture of Blondal, Hatami, Hatami, Lalov, and Tretiak. Hanner matrices give lower bounds of the same order, for square matrices and for explicit families with unequal logarithmic aspect ratios. The upper bound is obtained by constructing nested acyclic sets in the faces of a cube. A critical-point estimate supplies highly connected regions, weighted cell attachments keep the sets compatible across faces, and a cubical carrier theorem converts them into accurate finite covers and learning rules. Appending one constant coordinate turns list replicability into an exact geometric quantity, the face width, and graph-orientation classes satisfy exact product formulas. In particular, there are explicit classes with list replicability $4k+1$ and equivariant index $3k$, so the additive gap between the two is unbounded. We also develop entropy-tilting and two-sided Gibbs constructions, give combinatorial upper bounds adapted to structured matrices, compute exact values on classical families, and study the index of Paley matrices.
Read the full abstract

High Energy Physics (Phenomenology)

A comparative analysis of semileptonic $B_{(s)}\to D_{(s)}$ and $B_c\to(\eta_c,B,B_s)$ decays in the light-front quark model

We present a global analysis of the semileptonic transitions $B_{(s)}\to D_{(s)}$ and $B_c\to(\eta_c,B,B_s)$ by comparing five potential-model implementations of the QCD-motivated light-front quark model. We use the same fully Bakamjian–Thomas-based prescription for the weak current in all five models and keep their respective spectroscopic parameters fixed. We examine how confinement, radial mixing, and the treatment of hyperfine interactions jointly affect the form factors, differential decay rates, branching fractions, and lepton-flavor ratios. The internal Lorentz prefactors use the same free quark–antiquark kinematics as the on-shell constituent spinors in the weak-current kernels, while the physical meson masses determine the external kinematics. For each model and decay channel, the two independent extractions of $f_-(q^2)$ using the $(+,\perp)$ and $(+,-)$ current-component combinations agree within numerical accuracy. We perform a direct analytic continuation of the spacelike overlap integrals in the $q^+=0$ frame using a mass-weighted contour in complex transverse momentum space. This yields the form factors throughout the physical timelike region, including zero recoil, without parametrizing their $q^2$ dependence. Form factors and differential decay rates are presented as equal-weight model averages with pointwise minimum–maximum bands. The light-lepton-averaged ratios range from $0.282$ to $0.301$ for $R(D)$ and from $0.287$ to $0.303$ for $R(D_s)$.
Read the full abstract

Human-Computer Interaction

Early Signals of Problematic AI Use: The Longitudinal Roles of Motivation and Emotional Engagement

Longitudinal evidence about the psychological impact of conversational AI use remains scarce. We report a survey study of 114 adults who reported their level of problematic AI use one year after reporting their motivations (for escape, social, entertainment, and instrumental reasons) for engaging with AI. We found high levels of AI motivation of all types predicted later problematic AI use, especially when participants turned to AI for managing their feelings and overcoming difficulties. Our findings imply that AI motivation could be used as an early warning signal of potential problematic use, especially if individuals engage with AI for emotional purposes over time.
Read the full abstract

Information Retrieval

Yuanbao-Reranker: Document-Grounded Synthesis and Preference Distillation for Text Reranking

High-precision text reranking is a critical component in modern search, recommendation, and agentic retrieval-augmented generation (RAG) systems. We present Yuanbao-Reranker, a family of general-purpose dense rerankers spanning 0.6B, 4B, and 8B parameter scales built upon the Qwen3 backbone. To construct diverse and reliable ranking supervision at scale, we pair answer-first and query-type-guided document-grounded synthesis with a three-stage curation funnel that verifies positive support, eliminates semantic redundancy, and enforces balanced source-by-length quotas. For each curated pair, we screen real retrieval pools with a teacher score margin to filter out potential false negatives and sample Hard, Medium, and Moderate negative tiers, yielding 2.75M $1{:}3$ ranking groups. We train the models by jointly optimizing pointwise relevance cross-entropy and a bounded minimum-gap hinge objective over both positive–negative and negative–negative pairs. This formulation transfers fine-grained teacher preferences across candidate tiers without over-constraining well-separated predictions, supporting both cross-scale and same-scale distillation followed by layerwise adaptive model merging. Across five comprehensive benchmarks spanning Chinese (CMTEB-R), multilingual (MMTEB-R), English (BEIR), long-document (MLDR), and code (MTEB-Code) retrieval, Yuanbao-Reranker consistently leads every parameter tier: the 0.6B variant reaches 70.96 Avg. to surpass 1B–4B active-parameter baselines, the 4B variant reaches 73.63 Avg. to outperform the 8B teacher (72.68), and the 8B variant achieves a state-of-the-art 74.21 Avg., with substantial gains generalizing across distinct first-stage retrievers.
Read the full abstract

All authors Hongyuan Xu, Dezhi Ye, Jiahao Cao, Jie Liu, Haijin Liang, Haibo Shi, Jin Ma

Computation and Language

Coding or Cheating? An Investigation of Frontier Coding LLMs

Frontier agentic LLMs are trained on verifiable rewards in reinforcement learning (RL) environments built from real software repositories. A passing test suite is treated as the reward for solving the task. We show that, on hard problems, frontier models often satisfy these rewards by obtaining the answer instead of deriving it, known as reward hacking. Prior reports document this behavior on public benchmarks. We test its persistence on unseen problems. In a hardened environment where the answer cannot leak, nearly every model we tested attempted to recover the answer, escalating through tiers of increasingly specialized exploitation: mining the repository history for the upstream fix, inspecting the git object store directly, and sweeping the filesystem for a possible newer copy of the source or hidden tests. These attempts were concentrated on the hardest problems, and the models differed in whether, how often, and how deeply they escalated. Our position is that this behavior is a learned shortcut, acquired in RL environments that leak the answer, not a sign of intelligence. We release agent traces for these models to support further study at https://huggingface.co/datasets/logitly-ai/reward-hacking-raw-v1.
Read the full abstract

Cryptography and Security

QUASI: Post-Quantum Quorum Attestation of Pre-Validated Statements

We present QUASI, a newly conceived class of hash-based, post-quantum-secure protocols that enable a group of parties to produce a publicly verifiable attestation about a pre-validated statement. QUASI produces attestations that reduce the size and verification cost to a fraction of a full hash-based signatures collection, without relying on SNARKs or a trusted setup. QUASI rests on a surprising discovery: each party need not sign the entire statement. Authenticating only a small subset of symbols suffices for a quorum to attest to the whole. We turn this insight into a randomized construction that achieves reductions of more than 10x in per-party contribution size and hash computations for generation and verification, compared with full hash-based signatures. QUASI is a generic primitive supporting diverse applications. Operationally, QUASI requires no additional interactive setup or preprocessing beyond independent key generation by parties for the underlying hash-based authentication scheme. Two principal use cases are external attestation for signature-free consensus, enabling clients to verify statements about committed state; and preprocessing for SNARK-based aggregation, compounding QUASI's savings with succinct-proof compression.
Read the full abstract

Functional Analysis

Basis selection, universal sampling discretization, and sampling recovery

For every real $N$-dimensional subspace of $L_2$, we construct one orthonormal basis such that, for every $1\le v\le N$, at most $Cv\log(eN/v)$ positively weighted samples discretize the squared $L_2$ norm of all $v$-term combinations. Random rotations also give simultaneous weighted $L_p$ estimates and a sharp $C\log(eN)$ bound for the squared $L_2$ norm of the pointwise maximum. For prescribed coefficient classes, a change of measure yields sampling recovery estimates from $Cv\log^3(2v)$ samples under a lower Riesz inequality and a common square-integrable pointwise bound.
Read the full abstract

Statistical Methodology

Efficient Estimation and Consistent Class Number Selection for Semiparametric Class-Specific Index Regression

The study introduces a parsimonious semiparametric class-specific index regression that captures heterogeneous covariate–response relationships across latent subpopulations. In such a latent class model formulation, subject-specific indices are used as a complete-data representation of the realized latent labels. We propose a unified framework for estimating model parameters and recovering latent class memberships. For the class-specific index mean model, estimation and latent class recovery are conducted via a pseudo-least-squares criterion with a separation penalty, and a semiparametric information criterion is proposed for consistently estimating the number of classes. Exploiting the response distribution, we develop a Bayes classification rule that maximizes the asymptotic probability of correct assignment and a pseudo-maximum likelihood criterion that yields class index coefficient estimators with asymptotic semiparametric efficiency. This refinement step characterizes latent class membership probabilities via sufficient dimension reduction of the covariate space. Simulations and applications to marketing and biomedical datasets demonstrate superior finite-sample performance of the proposed methodology in parameter estimation, latent class recovery, and class number selection.
Read the full abstract

All authors Jen-Chieh Teng, Sheng-Hsin Fan, Chin-Tsang Chiang, Alvin Lim

Robotics

Detecting and Explaining Robot Policy Failures from Sparse Internal Features

Visuomotor policies can handle many robot manipulation tasks, but understanding their failures remains difficult. Unfamiliar conditions and small action errors can lead to failure, yet watching the robot alone tells us little about what went wrong inside the policy. We introduce MIND (Mechanistic Interpretability for Neural Diagnosis), a method that looks inside a trained policy to detect failures during execution and identify the internal concepts involved. MIND uses a sparse autoencoder to break down the policy's internal activations into features that can be studied individually. We give meaning to these features by linking their activity to the robot's position, movement, and surroundings, and by steering their activations to examine how they affect the robot's actions. From successful and failed executions, we learn how the features behave on the way to a failure. During execution, MIND flags deviations from these patterns and uses the meaningful features contributing most to the change to describe the problem. Experiments in simulation and on a physical robot show that changes in these internal features help detect and explain failures. By using the same features for detection and diagnosis, MIND connects failure signals to understandable changes inside the policy.
Read the full abstract

All authors Riad Ahmed, Sujosh Nag, Moniruzzaman Akash, Momotaz Begum

Artificial Intelligence

Too Big to Inspect? Ethically Auditing Large Datasets Through Small Samples

Big techs and researchers increasingly claim that their training datasets are too large to inspect for bias and ethical problems. However, in the race to build ever-large models, they have relied on automated web scraping and no human curation. Datasets built this way absorb the stereotypes, exclusionary norms, and historical prejudices of their sources, and the models trained on them can reproduce and amplify those patterns. Because manually auditing billions of samples is humanly impossible, we need practical, effective methods to identify biases in these datasets. This paper studies the feasibility of ethically auditing large datasets through small samples and proposes a practical method for identifying structural biases by applying AI Ethics Tools (AIETs) to random samples. We applied Datasheets for Datasets, Data Statements, Data Nutrition Labels, and Data Cards to a 1% sample of Framed Multi30K, and Datasheets for Datasets to two 0.00006% samples of relaion2B-en-research-safe. Our results show that evaluating small samples with AIETs can reveal ethical problems, structural biases, and technical inconsistencies. Although industry practice assumes that identifying problems requires analyzing the entire dataset, we show that a random subset supports a viable qualitative analysis without overwhelming computational or human resources. Demonstrating that a dataset is biased does not require examining every instance: if a bias appears in a random sample, it is a flaw of the creation process. The AIETs' questions acted as provocateurs, evoking different perspectives and analysis methods for the same dataset, but the analysis and its results depended fundamentally on the human factor. We hope these findings contribute to urgent public discussions about ethics, transparency, accountability, and AI regulation by offering a practical pathway for responsibly auditing massive datasets.
Read the full abstract

All authors Giovana Marchetti, Jhessica Silva, Tiago Torrent, Carlos Caetano, Sandra Avila

Databases

Measuring Tabular Data Formats Cold: A Replicated Cold-Cache Benchmark of Serialization, Scan, and Index Performance on Real Datasets

Choosing a file format for tabular data is a routine engineering decision, yet published comparisons often time reads served from the operating-system page cache, use one session, or rely on a single synthetic dataset. This article benchmarks fifteen format and access-path configurations (CSV, JSON Lines, Pickle, Feather, Parquet, ORC, DuckDB, and SQLite, with several access paths) on two real public datasets, New York City taxi trips and United States airline on-time records, at 50,000 to 3 million rows. Every measurement uses an explicitly cold cache, randomized execution order, and three independent sessions (with restarts on the laptop), and is analyzed with bootstrap intervals and Holm-adjusted permutation tests. Parquet answered a filtered aggregation 2.6 to 11 times faster than CSV, while JSON Lines was 10 to 27 times slower. A secondary index made SQLite queries 2.6 to 4.8 times slower on randomly ordered data and helped only on data sorted by the filtered column, at low selectivity. A row filter inside the Parquet reader was significantly faster than a projected scan on flight data from one million rows and on both datasets at three million, and row-group skipping needs several row groups, which the default writer produced only at 3 million rows. A 4 GB Linux host ran out of memory on configurations that a 24 GB laptop completed and reversed some orderings, such as Pickle versus Parquet. Code, derived data, and raw timings are released.
Read the full abstract

Machine Learning

Regularization Amplifies Subliminal Learning

Recent work has shown that hidden biases can be transferred through data with no apparent semantic relationship to those biases. Can regularization limit this covert influence? By constraining adaptation, regularization might be expected to suppress behavioral changes on seemingly unrelated inputs. We systematically study this hypothesis in subliminal learning, as well as in related settings of covert influence, including phantom transfer and logit-linear selection. We find the opposite: regularization amplifies the transfer of hidden biases. In particular, subliminal learning is typically absent under full fine-tuning, but low-rank adaptation or explicit penalties can induce strong transfer. Phantom transfer and logit-linear selection show similar trends. Adding more data, an implicit form of regularization, also strengthens covert influence. Together, these findings suggest that the simplicity biases induced by regularization may facilitate covert influence. We further hypothesize that the level of regularization that maximizes covert transfer depends on task complexity.
Read the full abstract

All authors Lawrence Feng, Jacob Mitchell Springer, Gaurav R. Ghosal, Ziqian Zhong, Aditi Raghunathan

Optimization and Control

Discrete-Time Zero-Sum Markov Games with Control, Stopping, and State-Dependent Discounting

We study a discrete-time two-player zero-sum stochastic game with control and stopping, where both players can stop the game, the discount factors depends on the states, and the running and terminal rewards may be unbounded. A contraction argument for the constrained Shapley equation establishes the existence of the value and a Nash equilibrium consisting of stationary controls and stopping times. We further provides algorithms for approximating the value and computing an $\varepsilon$-Nash equilibrium. A numerical example with state-dependent discount factors illustrates the results.
Read the full abstract

Cryptography and Security

The Limits of Information-Flow Control Enforcement for Agentic Systems

Information-flow control (IFC) is a promising approach to enforce policy in agentic AI systems. However, IFC suffers from label creep, a phenomenon where monotonically increasing security labels propagate through a system and progressively limit its utility. Label creep in agentic AI systems can be mitigated using two major strategies: (1) improved system design and (2) label declassification. In this paper, we characterize how far system design alone can mitigate label creep in agentic systems. Specifically, we measure how system topology, scale, and agent composition affect utility loss and security enforcement cost proxies for conservative IFC enforcement in a stateful agentic system simulator. We develop a pipeline to systematically vary these features and measure their impacts on label propagation. Our results demonstrate strong associations between structural graph features and our IFC utility loss metric (up to a Spearman correlation $|\rho|$ of $0.96$) and show that system size and tool allocation substantially affect how labels propagate. These findings identify design choices to reduce IFC utility loss, but they also demonstrate that system design alone cannot fully prevent utility loss due to monotonic label propagation. This motivates the development of label declassification mechanisms that aid these design choices and diminish IFC utility cost.
Read the full abstract

All authors Jonathan Gregory, Kyle Domico, Kunyang Li, Jean-Charles Noirot Ferrand, Patrick McDaniel

Programming Languages

Crustify: Repository-Level Generation of Safe Rust Bindings with LLM Agents and Coverage-Guided Testing

Wrapping C code in a safe Rust interface lets Rust applications use it without writing unsafe code, until the code itself is migrated to Rust. Such bindings are tedious and error-prone to write, however: their internals use unsafe code that the Rust compiler does not check, so a single mistake can introduce undefined behavior (UB), and they have so far been written mostly by hand. We present Crustify, an agentic co-translator that generates safe Rust bindings for repository-level C libraries. Crustify schedules a target's types and symbols into dependency-ordered waves of independent batches that agents bind concurrently. Its agents derive the ownership and lifecycle contract of every pointer and type they wrap, check each binding with coverage-guided equivalence and UB tests under sanitizers and Miri, and review every wave. They build on three tools that we contribute: (a) ffibox, Rust primitives for C/Rust interop that mirror Box, Arc, and Vec for C objects without ever forming a reference to memory that C may alias; (b) Wavefront, a neuro-symbolic static analyzer that orders a target's items by their dependencies, including those that casts and type-erased pointers hide; and (c) ffimap, a rustc-based analysis that maps the unsafe surface of bindings. On libgit2 and ffmpeg, with one developer in the loop, Crustify generated safe bindings for 41.9% and 42.8% of their public APIs in about 51.9 and 58.9 hours. On the interface consumers are expected to use, only 1.6% and 0.4% of the bindings' public functions are unsafe, and only 0.1% and 0.6% have a raw pointer in their signature. Their tests cover 98.92% and 99.07% of the Rust code and 45.95% and 67.20% of libgit2 and of ffmpeg's libavutil, and the reviewers fixed 8 and 13 defects that caused UB reachable from Rust. To our knowledge, this is the first work to automate the generation of safe Rust bindings for C libraries.
Read the full abstract

All authors Marius Momeu, Marius Arvinte, Cory Cornelius, Koushik Sen