Artificial Intelligence
Context-dependent agent evaluation with orthogonal equilibrium learning
Haorui Ma, Zehua Zang, Jiangmeng Li et al. (+3 more)
Many applications require to evaluate
agents under contextual information (e.g., a
prompt, task, or user group). We study how to perform such context-dependent
agent evaluation from offline feedback. Existing score-based models for this purpose (e.g., Bradley-Terry) impose a transitive preference ordering, which fails to reflect collective preferences when human judgements are heterogeneous. Inspired by social choice theory, we frame evaluation as a contextual game between two players, each selecting a distribution over
agents as the strategy to receive greater collective preference than the other. Then, the support of the Nash
equilibrium defines a context-specific set of winners. However, learning context-specific
equilibria from offline logs is difficult because each context reveals human feedback on only a subset of
agents, and, hence, a naive plug-in
estimator can therefore be biased. To address these challenges, we propose NashEval, a general framework for robust contextual
equilibrium learning. NashEval first constructs debiased estimates of the contextual payoff
matrix that characterizes the game. NashEval then learns the context-to-
equilibrium mapping with a tailored orthogonal loss, which avoids the need to solve a separate game for each context. We show theoretically that errors in estimating the nuisance functions underlying the payoff
matrix affect the risk of the learned
equilibrium (i.e., exploitability) only through higher-order terms. Across various experiments, NashEval improves robustness of
equilibrium learning and consistently identifies the set of top-performing
agents across contexts.
Read the full abstract ▾
All authors Haorui Ma, Zehua Zang, Jiangmeng Li, Yi Li, Fanjing Xu, Stefan Feuerriegel
Computer Vision
Auditing Quality Filters for Long-Tail Human Data Curation
Rishav Agarwal, Nirshal Chandra Sekar and Anirudh Vemula
Robots on construction sites must detect workers who are kneeling or bending, which we call low poses. These workers can be lost from training
datasets during automatic labeling. We study a pipeline that detects people, estimates their body joints using NLF, and groups similar poses. Low poses account for only about 2 percent of the retained examples. This low share may partly reflect the pipeline's quality filter, which rejects examples with low detection confidence or uncertain joint estimates. We examine this filtering using four alternative pose clues: bounding-box shape, vertical body span, pose grouping aligned to the scene's vertical direction, and image appearance. All four suggest that low poses are rejected by the filter more often. Separately, controlled simulated scenes show that a person detector fine-tuned on a public construction
dataset misses more workers in these poses even when we correct their bounding box height is matched to that of standing workers. These findings suggest that low poses are scarce and hard to find, and we cannot rely on bounding boxes or poses for long-tail human data curation.
Read the full abstract ▾
Materials Science
The Young-E(3) Tensor Product Decomposition for Rotation and Permutation Equivariant Cluster Expansions
James M. Goff and Aidan P. Thompson
A generalization of fixed-
lattice cluster expansions (CE) and atomic cluster expansions (ACE) is presented that expresses both rotation and permutation
symmetries. By performing Schur–Weyl decomposition in Young subgroup-stabilized carriers, followed by joint coupling of rotation and permutation representations, arbitrary $S_N\times SO(3)$ tensor product carriers are generated. This allows us to resolve complete, orthonormal joint rotation and permutation-adapted bases for arbitrary tensor product ranks and arbitrary angular and radial tensor product content. We show that this very general Young–E(3) (YE3T) tensor product basis contains the ACE basis as a subset, corresponding to the special case where permutation
symmetry character is restricted to the fully symmetric carrier. Rather than constructing an overcomplete rotation and permutation-invariant basis and reducing it a posteriori, Barthelemy et al. recently demonstrated scaling benefits by not constructing an overcomplete basis. YE3T directly produces a complete orthonormal basis without an overcomplete step and rigorously extends to permutation characters beyond permutation-symmetric features. The YE3T decomposition yields new joint irreducible rotation- and permutation-equivariant basis sets that surpass existing rotation-adapted expansions in both speed and accuracy, defining a new Pareto front for machine-learned interatomic potentials. We show that the tunability of the permutation
symmetry character makes the basis useful for both atomistic and electronic-structure simulations.
Read the full abstract ▾
Solar and Stellar Astrophysics
Spectroscopic Signatures of Nonuniform Lateral Expansion of a Coronal Mass Ejection in the Low Corona with ADITYA-L1/VELC
Saurabh Tripathi, Tanmoy Samanta and Jayant Joshi
Coronal mass ejections (CMEs) undergo rapid
evolution in the low corona ($<2R_\odot$), yet their three-dimensional kinematics remain poorly constrained. We present a spectroscopic and imaging analysis of a CME observed on 2024 August 5 using the Visible Emission Line Coronagraph (VELC) on board ADITYA-L1 and Atmospheric Imaging Assembly (AIA)/
Solar Dynamics Observator (SDO). The VELC Fe XIV 5303 Å spectra captured the erupting CME as it crossed the western slit. Space-time maps from single-Gaussian fits reveal coronal dimming and enhanced Doppler velocities during the eruption. Detailed spectral analysis reveals strongly Doppler-shifted secondary components associated with the CME front. These components appear as two spatially separated emission branches on opposite sides of the rest wavelength, producing a distinct bifurcated spectral morphology superposed on the nearly stationary reduced-intensity coronal emission. Double-Gaussian fitting yields line-of-sight (LOS)
plasma velocities of up to 218 km/s, likely a lower limit as the secondary component reaches the edge of spectral window. AIA 211 Å observations measure a projected radial speed of 1135 km/s, larger than the plane-of-sky (POS) lateral speed of 692 km/s along the slit. To interpret the observed CME
evolution and spectral structure, we developed a simplified three-dimensional
geometric-kinematic model of an expanding axially tilted CME shell. Synthetic spectra from the model show that an inclined CME undergoing asymmetric lateral expansion can produce a bifurcated spectral morphology similar to that observed, with velocity components projected in the radial, POS-lateral, and LOS-lateral directions. Our results demonstrate that combining spectroscopy and EUV imaging constrains different components of CME kinematics during early
evolution.
Read the full abstract ▾
Machine Learning
CyberWorld: World Models for Sample-Efficient Autonomous Cyber Defense
Ryozo Masukawa, Sanggeon Yun, Raheeb Hassan et al. (+3 more)
Deep reinforcement learning has become a prominent approach to autonomous cyber defense. Existing methods are predominantly model-free and consequently require extensive environment interaction. World models provide an alternative by learning predictive dynamics and optimizing policies through imagined trajectories, yielding substantial gains in sample efficiency in robotics and
embodied control. Extending this paradigm to cybersecurity raises a fundamental question: what should constitute the "world" in a cyber world model? We introduce CyberWorld, a Dreamer-style world modeling framework that learns latent cyber dynamics from vector, graph, textual, and
multimodal representations of the defended network. Across all four scoreable CyberWheel attack strategies, the graph-based CyberWorld variant exceeds a strategy-agnostic control after 3.6k-15.8k environment steps, compared with millions of steps required by model-free PPO. Across representation choices, graph structure provides greater robustness under
topology-dependent attacks, while simpler representations remain competitive in overall performance. Among successful runs, the number of episodes required to reach the control remains approximately constant as network size increases from 15 to 100 hosts. These results establish learned cyber dynamics as a sample-efficient and scalable basis for autonomous defense, and identify world representation as a central design axis for robustness and scalability.
Read the full abstract ▾
All authors Ryozo Masukawa, Sanggeon Yun, Raheeb Hassan, Hyunwoo Oh, SungHeon Jeong, Mohsen Imani
Sound
NVAlign: Direct-Gradient Optimization for Non-Verbal Control in Continuous Autoregressive Flow Matching Text-to-Speech
Qiaolin Wang, Pedro Sandoval-Segura, Anunaya Joshi et al. (+2 more)
While modern text-to-
speech (TTS) systems generate highly natural
speech and support inline non-verbal vocalization (NVV) tags, accurate control over these events remains challenging. A key gap is the lack of established post-training methods for non-verbal control in continuous autoregressive flow-matching TTS. To this end, we present NVAlign, a direct-gradient post-training framework for NVV tag-following in this architecture. We first perform supervised fine-tuning (SFT) of TTS models and an NVV-aware automatic
speech recognition (NV-ASR) model on NVV-annotated
speech, then freeze the NV-ASR model to serve as the reward model for post-training. A two-step gradient surrogate enables efficient reward backpropagation through the flow-matching sampler to jointly update the autoregressive backbone and acoustic flow head. Fidelity penalties and reference-velocity regularization help preserve speaker similarity and
speech quality. Results from NVV-SuperBench and human listening evaluations show that NVAlign improves tag-following accuracy over SFT and Flow-GRPO baselines. These findings demonstrate that direct reward-gradient optimization can improve non-verbal control in continuous autoregressive flow-matching TTS. Audio samples are available at https://nvalign.github.io/.
Read the full abstract ▾
All authors Qiaolin Wang, Pedro Sandoval-Segura, Anunaya Joshi, Edvardas Jurkonis, Jake Downie
Accelerator Physics
Incorporating Heightened Scrutiny into a Large HEP Software Project
Philippe Canal, Chris Green, Kyle Knoepfel et al. (+5 more)
In 2023, DUNE began re-evaluating the requirements of its data-processing framework, which led to commissioning a new design that would better fit
neutrino physics than the existing reconstruction frameworks designed for
collider physics. Due to the radical changes expected, significant multi-institutional effort has been directed toward the creation of the Phlex framework. In addition, the tight timelines in which to implement such a framework have invited scrutiny from various parties, including DUNE itself, the host laboratory Fermilab, and the US Department of Energy. After briefly discussing the unique needs of
neutrino physics, we will recount how the Phlex development team has approached the design and development of a framework in the context of this heightened scrutiny. This process began with a systems-engineering approach to formally manage the requirements DUNE has of its framework. What followed was, for the first time in the HEP community, a review of the framework's conceptual design by a panel of external framework experts before developers proceeded to the implementation. The implementation efforts have led to a series of prototypes that are being used to provide rapid feedback from Phlex users to framework developers. We will discuss how each of these steps has resulted in a strong design that has the backing of the DUNE experiment, Fermilab, and the US Department of Energy.
Read the full abstract ▾
All authors Philippe Canal, Chris Green, Kyle Knoepfel, Wim Lavrijsen, Adam Lyon, Marc Paterno, Saba Sehrish, Beojan Stanislaus
Machine Learning
FARE: Deep Reinforcement Learning For Fair Exposure Constrained Uncertainty Aware Financial Content Personalization
Arundeep Chinta, Lucas Vinh Tran and Jay Katukuri
Content personalization systems in financial services must ensure fair exposure across diverse offerings-a requirement driven by contractual obligations and the need to prevent "rich-get-richer" dynamics where content with high click-through rate (CTR) dominates while other relevant products receive minimal visibility. Share of Voice (SOV) constraints, which guarantee each content category a target fraction of top-position exposure, address this by promoting product diversity and balanced user discovery. While re-ranking layers atop CTR models are common in practice, we propose two key novelties: (1) framing SOV-constrained ranking as a deep reinforcement learning problem analogous to constrained trade execution in algorithmic finance, and (2) explicitly incorporating CTR prediction uncertainty into the
agent's state space and policy design-enabling larger ranking adjustments for high-uncertainty predictions where deviation from CTR-optimal ordering is less costly. We introduce FARE (Fair Ranking Executor), a modular uncertainty-aware execution layer that translates any black-box CTR model's predictions into SOV-fair rankings without retraining the underlying model. Our uncertainty-weighted proportional control policy (FARE-PC) and learned neural policies (FARE-ES, FARE-PPO) demonstrate that uncertainty-aware approaches can substantially reduce SOV deviation from
fairness targets while minimizing engagement loss, with gradient-free
evolution strategies outperforming policy gradient methods on synthetic data and the ordering reversing on KuaiRand-Pure.
Read the full abstract ▾
Quantum Physics
Scattering amplitudes from quantum hardware a la RESOs
Raul A Briceno, Ivan M Burbano, Anthony N Ciavarella et al. (+3 more)
We report the first
quantum-hardware implementation of Real-time
Estimators for Scattering Observables (RESOs). The physical model is a one-dimensional
lattice theory of non-relativistic
spin-1/2
fermions with a single contact interaction. The
lattice sizes studied are up to $N_x=54$
lattice sites and up to $20$ time steps. Spacetime correlation functions were computed on IBM's
quantum computers, using $109$
qubits with $10,806$ CZ gates applied, reaching a two-
qubit gate depth of $246$. We demonstrate how to study bound states and extract scattering amplitudes from these correlation functions.
Read the full abstract ▾
All authors Raul A Briceno, Ivan M Burbano, Anthony N Ciavarella, Ermal Rrapaj, Thomas R Richardson, Andre Walker-Loud
Materials Science
Knowledge-Driven XRD Phase Identification via Multi-View Retrieval and Explanation
Doaa Mohamed and Markus Stricker
X-ray diffraction (XRD) is a experimental technique for determining the phase composition and structure of crystalline materials. However, interpreting XRD patterns is challenging, particularly in high-throughput materials discovery, where many novel materials may need to be characterized and no reference patterns are available. Consequently, machine learning is increasingly used to accelerate and automate the analysis while reducing errors associated with human interpretation. We propose a multi-decision framework for XRD phase analysis that integrates representation learning, similarity-based
retrieval, and explainable decision support within a unified reference database. A convolutional autoencoder learns compact latent representations of XRD patterns that preserve structural similarity while remaining robust to variations arising from experimental noise and measurement conditions. By integrating multiple decision pathways within a shared latent space, the framework moves beyond single-label prediction toward ranked and interpretable phase analysis that mirrors expert practice. During inference, complementary decision mechanisms are applied, including latent-space classification and
retrieval, explanation-guided similarity using Integrated Gradients, and composition-based similarity search. These mechanisms generate ranked candidate phase lists that are aggregated into a final prediction with an associated confidence score. Experiments on synthetic
datasets demonstrate strong predictive performance, achieving 98.85\,% accuracy for
crystal system classification and 95.82\,% accuracy for space group prediction on the test set, while maintaining robustness under realistic perturbations. The framework supports reliable, analyst-friendly identification of
crystal phases and structures in high-throughput and exploratory materials discovery settings.
Read the full abstract ▾
Programming Languages
Efficient Extraction for Effectful E-Graphs
Oliver Flatt, Anjali Pal, Yihong Zhang et al. (+7 more)
Egraphs have enabled recent advances in program optimization, synthesis, and verification, yet remain difficult to apply to effectful programs whose
memory and I/O operations must respect execution order. Existing effect-aware extraction
algorithms rely on integer linear programming (ILP) and dominate total runtime. We introduce Statewalk DP, a new extraction
algorithm that enforces effect ordering efficiently without external solvers. We prove that finding any effect-safe extraction is NP-complete, but show that Statewalk DP is tractable in statewalk width, a parameter that measures the complexity of dataflow interactions among effects. In practice, statewalk width generally remains small, enabling Statewalk DP to achieve order-of-magnitude speedups over ILP extraction while producing programs comparable to LLVM across our
benchmarks. We implement the
algorithm in eggcc, a prototype egraph-based compiler for imperative Bril programs and demonstrate that effect-aware extraction is no longer a bottleneck.
Read the full abstract ▾
All authors Oliver Flatt, Anjali Pal, Yihong Zhang, Ryan Tjoa, Kirsten Graham, Alex Fischman, Chandrakana Nandi, Eli Rosenthal, Zachary Tatlock, Haobin Ni
Solar and Stellar Astrophysics
Negative and Positive Cascade Rates of Slow Alfvénic Turbulence in Switchback and Non-switchback Intervals: Wind Observations
Sofiane Bourouaine
We estimate the turbulent cascade rate in imbalanced Alfvénic
turbulence in the slow
solar wind using a Politano–Pouquet-based approach that does not require an isotropic assumption. We analyze selected 8-hour intervals provided by Wind spacecraft. These intervals are characterized by near-homogeneity, weak velocity shear, and imbalanced Alfvénic fluctuations, and are classified as switchback or non-switchback depending on the angular spread of the
magnetic field relative to the mean field. Switchback intervals are identified by large
magnetic-field rotations, whereas non-switchback intervals exhibit only small angular deviations. Our results show that non-switchback intervals exhibit a positive total cascade rate and power spectra consistent with a $-3/2$ scaling, indicative of a forward energy cascade. In contrast, switchback intervals are characterized by a negative total cascade rate (inverse cascade), and by steeper power spectra with a spectral index of approximately $-1.6$. This indicates that large
magnetic-field rotations can substantially modify the scale-to-scale energy transfer in Alfvénic slow
solar wind. The results suggest that switchbacks may be associated with additional physical processes, such as local instabilities, or reconnection-related structures, that alter the direction and magnitude of the turbulent cascade. These findings may help explain previous reports of negative cascade rates in imbalanced Alfvénic
solar-wind
turbulence.
Read the full abstract ▾
Robotics
Differentiable Dynamics for Autonomous Micro-Mobility Navigation
Grace Cai, Joey Lee, Nithin Parepally et al. (+2 more)
Autonomous micro-mobility vehicles (MMVs) such as wheelchairs, scooters, and bicycles have the potential to improve mobility access and support safe low-speed transportation in pedestrian-shared spaces. Achieving MMV autonomy will require realistic, predictable MMV motion. However, many existing autonomous vehicle stacks rely on simplified kinematic models that fail to capture key MMV characteristics such as tire slip, friction, and wheel layouts, limiting realism and gradient-based optimization. In this paper, we explore differentiable formulations of dynamics models for autonomous micro-mobility systems. We first construct DiffKBM, a differentiable version of the kinematic bicycle model (KBM). Then, we introduce DiffGM3, a differentiable formulation of the General Micro-Mobility Model (GM3), a unified tire-based dynamics formulation for micro-mobility vehicles that supports a wide range of MMV configurations. DiffKBM and DiffGM3 enable end-to-end differentiable optimization through MMV dynamics, making them suitable for integration into differentiable autonomy stacks. We evaluate these dynamics models in both open-loop and closed-loop settings: (1) open-loop trajectory matching, where DiffKBM and DiffGM3 are integrated as a dynamics layer within DiffStack and optimized to reproduce real-world MMV trajectories, and (2) closed-loop autonomous
navigation, where DiffKBM and DiffGM3 are paired with a differentiable MPC controller in CrowdNav pedestrian scenarios. In the open-loop setting, DiffGM3 outperforms DiffKBM in reproducing trajectories with improvements in ADE and NLL across bicycle, scooter, and motorcycle modes, and reductions in planning loss for bicycle and motorcycle trajectories. We also find that, in closed-loop settings, DiffGM3 improves on DiffKBM's CrowdNav performance by producing 55% fewer collisions and a 75% lower discomfort frequency for the bicycle mode.
Read the full abstract ▾
All authors Grace Cai, Joey Lee, Nithin Parepally, Laura Zheng, Ming C. Lin
High Energy Physics (Theory)
Transient Coherence and Multiscale Spectral Structure after a Mass Quench in a 1+1-Dimensional Scalar Field
Gabriel Abellán and Nelson Bolívar
A mass quench generates anomalous pair correlations that are absent from the final particle occupations. For a free real scalar field in $1+1$ dimensions, we define a bilocal observable by smearing the statistical Keldysh correlator with a Morlet kernel on the finite post-quench relative-time interval. Its excess over the dephased correlator gives a dimensionless relative coherence weight. At fixed momentum, this weight is exactly independent of the initial temperature for Gaussian thermal states evolved with the same free
Hamiltonian: the common thermal amplitude cancels between numerator and denominator. The cancellation is not a purity criterion and is not exclusive to wavelets. We derive the finite-window response and show that the anomalous signal depends essentially on the observation window; an admissible zero-mean wavelet annihilates the relative-time-constant anomalous term on the full line. We compare this response with the signed finite-time Wigner spectrum and derive a distinct Gaussian-extension positivity bound. Exact smooth-quench occupations provide an independent dynamical reference. Finally, we formulate a renormalized Hartree extension and delimit the oscillatory and nonoscillatory regimes of a scalar mode on a fixed two-dimensional de Sitter background.
Read the full abstract ▾
Operator Algebras
Orbit-breaking in Deaconu-Renault groupoids and models for UCT Kirchberg algebras
Robin J. Deeley, Ian F. Putnam and Karen R. Strung
We generalize the concept of orbit-breaking from the case of a homeomorphism to the case of a surjective local homeomorphism. This method creates open subgroupoids of the Deaconu–Renualt groupoid associated to the surjective local homeomorphism. When the map satisfies natural assumptions (including being locally expanding) the resulting groupoid $\mathrm{C}^*$-
algebras are unital UCT Kirchberg
algebras. A specific class of Deaconu–Renault systems are constructed, each modelling the Cuntz
algebra $\mathcal{O}_2$. The $\mathrm{C}^*$-
algebras obtained from orbit-breaking for such systems come with explicit
embeddings into $\mathcal{O}_2$. We also introduce a method for orbit-breaking after removing a fixed point. Using this method, we can realize all stable UCT Kirchberg
algebras with each coming with an
embeddings into $\mathcal O_2\otimes\mathcal K$. Moreover, we show that, given a homomorphism between countable graded abelian groups, there is a $^*$-homomorphism between orbit-breaking
algebras obtained from a groupoid inclusion. In particular, we obtain explicit $^*$-homomorphisms between stable UCT Kirchberg
algebras from any group homomorphism between their $K$-groups.
Read the full abstract ▾
Machine Learning
DOHF: Online Diffusion Fine-tuning with Doob's $h$-transform Guidance
Zhengyi Guo, Jiayuan Sheng, Wenpin Tang et al. (+1 more)
Reward-based
diffusion fine-tuning faces practical challenges when desirable outcomes are rare or conditioning corrections are costly to estimate. In this work, we propose
Diffusion Online $h$-guidance Fine-tuning (DOHF), which turns Doob's $h$-transform into a practical online training
algorithm. DOHF assigns optimality weights to generated samples, estimates the normalized local correction $\nabla\log h$ under the current rollout policy, and distills it directly into the generative model. Theoretically, we characterize the population-optimal DiffusionNFT update as well as the various classfier free guidance methods through a unified $h$-transform perspective. Methodologically, our framework accommodates black-box and non-differentiable rewards without additional network evaluations. We further show improved
alignments under three empirical scenarios. Our work demonstrates how adapting probabilistic conditioning through inexpensive estimation and iterative
distillation can improve generative learning across statistical sampling and visual generation.
Read the full abstract ▾
All authors Zhengyi Guo, Jiayuan Sheng, Wenpin Tang, David D. Yao
Machine Learning
TemporalGraphLLM: Temporal Graph Neural Networks with Large Language Models for Dynamic Text-Attributed Graphs
Moran Beladev, Or Eitan, Gilad Katz et al. (+1 more)
Dynamic text-attributed graphs (DTAGs), where nodes, edges, and textual attributes evolve over time, are crucial in applications such as social networks, citation graphs, and knowledge graphs. However, existing approaches struggle to jointly model the temporal
evolution of graph structures and the semantic richness of textual attributes. While Temporal Graph Neural Networks (TGNNs) capture evolving node relationships, they often lack contextual text
reasoning. Conversely, Large
Language Models (LLMs) excel in textual understanding but struggle with structured graph
reasoning in temporal settings. To bridge this gap, we propose TemporalGraphLLM, a novel framework that can integrate any temporal GNN with an LLM for enhanced
reasoning in DTAGs. Our approach fine-tunes LLMs using graph-time-aware instruction tuning and novel temporal GNNs injection to replace dedicated added
tokens with graph
embeddings. TemporalGraphLLM effectively leverages pretrained TGNNs within an LLM framework to achieve state-of-the-art performance on edge classification, link prediction, and edge-based text generation tasks. Extensive evaluation on real-world dynamic graph
datasets demonstrates state-of-the-art performance. Our findings highlight the synergistic potential of LLMs and TGNNs, opening new directions for learning on evolving graphs.
Read the full abstract ▾
All authors Moran Beladev, Or Eitan, Gilad Katz, Lior Rokach
Robotics
Efficient Bezier Velocity Optimization for Free-Floating Space Manipulators
Duo Zhang, Zhizhuo Zhang, Xiaoli Bai et al. (+1 more)
We present FAVOR (Free-floating Arm Velocity Optimization with Recursive Sensitivities), a planner for collision-free reaching, tracking, and prescribed-time pre-grasp interception on an unactuated spacecraft. It optimizes Bezier joint-velocity curves with linear velocity, acceleration, and continuity constraints. Decision dimension is independent of rollout resolution. Analytical recursive sensitivities provide task and clearance gradients through the coupled base-arm motion. Parallel evaluation, caching, and incremental collision discovery reduce computation. With a seven-DoF arm and five simulated spacecraft models, FAVOR achieves 99.8% point-to-point success with 1.779 s mean computation, versus 62.9% and 50.270 s for an IK-initialized position-spline baseline. Means include failures and timeouts. FAVOR completes 30 of 36 tracking cases, versus 15 for single-step QP, and all 36 interception instances, versus 22 for the spline baseline. A controlled ablation shows that finite differences increase mean planning time 4.7-fold.
Read the full abstract ▾
All authors Duo Zhang, Zhizhuo Zhang, Xiaoli Bai, Jingjin Yu
High Energy Astrophysical Phenomena
Not all spins of black-hole binaries formed in clusters are isotropic and not all spins of black-hole binaries formed in isolation are aligned
Sofia Dossena, Davide Gerosa and Tristan Bruel
Spin directions are promising observables for distinguishing between the formation channels of merging
stellar-mass binary black holes with gravitational-wave observations. In this work, we challenge the standard expectation that binaries formed via dynamical interactions in
stellar clusters have isotropically distributed
spins and that binaries formed in isolation have
spins aligned with the orbital angular momentum. For dynamically formed binaries, we account for observational findings suggesting that the
spins of
stars in clusters may exhibit a preferred degree of
alignment and construct a simple geometrical description of the resulting black-hole
spin directions. This results in an analytical joint distribution for the polar angles of first-generation black-hole binaries, whose key feature is a correlation between the two
spin-orbit tilts. We show that, under reasonable assumptions and independently of the initial
spin configuration,
spins are fully randomized after a single black-hole merger, implying that binaries formed through hierarchical mergers have isotropically distributed
spin orientations. For isolated binaries, we introduce a simple model for
spin evolution as a function of the natal kick imparted during the first
supernova explosion, while accounting for
tidal interactions and mass-ratio reversal. The resulting distribution is also fully analytic. We investigate the impact of relativistic
spin precession, as well as that of the finite escape speed of clusters, on our geometrically derived distributions. Our results provide a simple analytical framework for modeling the
spin directions of merging black holes in gravitational-wave population fits.
Read the full abstract ▾
Cryptography and Security
TokenScanner: Detecting Backdoors and Discovering Triggers in Text-to-Image LoRAs via Full Vocabulary Scanning
Boliang Liu and Jing Zhang
LoRAs are widely studied for adapting base text-to-image
diffusion models. However, a backdoored LoRA can hide a
backdoor: it behaves normally in most cases, but produces attacker-specified content (the
backdoor target) when a hidden
backdoor trigger appears in the input
prompt. Detecting such
backdoors before using an untrusted LoRA is important for the safety of LoRA adaptation. We present TokenScanner, a model-level vocabulary scanner for
backdoor detection within a LoRA fine-tuned text-to-image
diffusion model, aiming to discover the malicious trigger for trustworthy LoRA adaptation. The key observation is that
backdoor trigger
tokens that appear in a larger proportion of training
prompts tend to induce more prominent
token-specific LoRA responses than those induced by unrelated
tokens. TokenScanner therefore scans the tokenizer vocabulary and measures
token-wise LoRA responses in the U-Net and the text encoder. It uses these responses to detect backdoored LoRAs and rank candidate trigger
tokens for subsequent testing of
backdoor activation. Experiments on seven
backdoor settings, comprising 840 backdoored LoRAs and 840 real-world benign test LoRAs, show that TokenScanner achieves 95.96% AUC and 96.55% TPR at an FPR of 10.95%. It also achieves 89.40% Hit@1 and 99.40% Hit@5 for trigger discovery across all seven settings.
Read the full abstract ▾
Cryptography and Security
A Large-Scale Benchmark and Risk Assessment of Traffic Analysis Attacks on Cloud LLM Services
Shahrooz Pouryousef, Jesus Lopez, Saeefa Rubaiyet Nowmi et al. (+4 more)
Cloud-based Large
language model (LLM) services create a network-level traffic side channel that can expose model,
prompt, and task behavior despite encryption. From packet sizes, directions, timing, and burst structure alone, a passive local observer can infer the serving model, the user's
prompt category, and the task executed by a collaborative multi-
agent system. Yet current evidence is fragmented across separate
datasets and settings, limiting reproducibility and comparison. We present, to our knowledge, the first unified measurement study and public
benchmark of encrypted LLM traffic across both user–LLM and multi-
agent executions. The large-scale
benchmark contains 60,000 user–LLM interactions across 10 models and 6
prompt categories, plus 2,838 multi-
agent executions covering 10 task categories and two coordination
topologies. Using only encrypted packet metadata, we assess the risk of traffic analysis attack by characterizing traffic signatures, identifying the features most associated with leakage, and testing robustness under
prompt reformulation, decoding-temperature changes, larger candidate model sets, and partial traffic observation. Model fingerprinting achieves 97.7% balanced accuracy,
prompt-category fingerprinting reaches 76.7% mean accuracy, and multi-
agent task fingerprinting achieves up to 90.7% accuracy.
Prompt reformulation weakens but does not remove model-specific leakage, and task fingerprints remain detectable even from a single
agent's traffic.
Read the full abstract ▾
All authors Shahrooz Pouryousef, Jesus Lopez, Saeefa Rubaiyet Nowmi, Md Mahmuduzzaman Kamol, Moinul Hossain, Muoi Tran, Mohammad Saidur Rahman
Computation and Language
Agents Can Use Base Models to Evade AI Detection
Bhuwan Dhingra and Danish Pruthi
We show that coding
agents equipped with a base
language model can successfully assemble responses from its samples to evade detection. Base models have been shown to evade commercial detectors, however, prior "humanization" techniques rely on using these models to paraphrase AI outputs over several iterations, which invariably results in semantic drift. In contrast, equipping coding
agents to directly orchestrate the writing process by stitching text samples from a base model allows it to produce outputs that are coherent, task-specific and generally high quality. We find that Claude Opus 5 operating in a Claude Code harness effectively orchestrates a local 32B parameter OLMo-2 base LM and sacrifices little task accuracy across
benchmarks spanning creative writing, factual grounding, health QA and instruction following, while using up to 90% base LM
tokens. Responses constructed in this manner reduce the effectiveness of both post-hoc detectors (Pangram v4 detection rate drops from 77% to 24%) and soft watermarking applied a priori to the
agent's generations (down to a simulated 10% detection at low FPR). While effective, this evasion requires a significantly larger number of input and output
tokens from the
agent, increasing the dollar cost per query up to 30x at API-pricing. Overall, this work demonstrates the effectiveness of a new class of
adversarial attacks against AI text detection, and urges post-hoc detection providers to include outputs of base models in their training.
Read the full abstract ▾
Chemical Physics
Heteroatom Position Controls Ultrafast Photodynamics of Oxazole and Isoxazole
Briony Downes-Ward, Paul Javed, Huynh V. S. Lam et al. (+27 more)
The ultrafast photochemistry of heterocyclic compounds is central to photobiology and materials chemistry, yet direct experimental observation of their structural dynamics remains rare. Here we present the first real-time structural characterization of photoinduced ring opening and subsequent fragmentation in the isomeric pair oxazole and isoxazole using MeV ultrafast electron diffraction (UED), complemented by non-adiabatic molecular dynamics simulations. Upon photoexcitation at 200 nm, both isomers undergo ring opening followed by fragmentation into various products, but with strikingly different dynamics governed by heteroatom positioning. Trajectory Surface Hopping simulations reproduce the experimental diffraction signatures, which are surprisingly similar for both isomers, and reveal distinct mechanistic pathways for the two isomers. For isoxazole, all trajectories exclusively undergo N-O bond cleavage within 40 fs, followed by sequential fragmentation into HCN + ketene and HCO + vinyl nitrene channels on the hundreds-of-femtoseconds timescale. Oxazole, by contrast, shows significantly slower ring opening (290 fs) with only 85% efficiency, proceeding primarily through O-C cleavage and accessing a richer landscape of intermediates including nitrile ylide and O-pyramidalized structures. Simulated diffraction patterns derived from trajectory ensembles, convolved with the experimental instrument response function, are in agreement with the branching ratios observed by UED. This synergy between UED and trajectory surface hopping provides an atomistic picture of how the simple interchange of heteroatom connectivity in structural isomers fundamentally reshapes excited-state potential surfaces, conical intersection accessibility, and photochemical outcome.
Read the full abstract ▾
All authors Briony Downes-Ward, Paul Javed, Huynh V. S. Lam, Sajib K. Saha, Surjendu Bhattacharyya, Martin Centurion, Xinxin Cheng, R. Joel England, Casey Foley, Smita Ganguly, Patrick L. Kramer, Jinxin Lang, Randy Lemons, Ming-Fu Lin, Yusong Liu, Tu T. Nguyen, Joao Pedro Figueira Nunes, Chatura Perera, Alexander H. Reid, Ethan Ross, Artem Rudenko, Xiaozhe Shen, John Searles, Anbu Selvam Venkatachalam, Enliang Wang, Stephen P. Weathersby, Vinod Kumarappan, Arthur G. Suits, Christine Aikens, Daniel Rolles
Artificial Intelligence
COUNTERMEM: World-Model Verified Counter-Factual Memory for Language Agents
Hongji Pu, Ruixiang Tang and Yongfeng Zhang
Existing
agent memory frameworks mainly create
memory through an
agent's interaction with the factual world, e.g., remembering feedback from actions taken to improve performance on future tasks. However, these frameworks seldom ask the "what if" question during
memory construction: what if a different action had been taken, would the feedback have changed, and how could this feedback become useful
memory? Obtaining such feedback directly in an active environment can be expensive and can alter the state needed for comparison. In this work, we introduce COUNTERMEM, a reinforcement-learning framework for constructing and using verified counterfactual
memory across tasks. After a failed action, COUNTERMEM evaluates local alternatives from a copy or reset of the original state using executable world models, such as tests, proof checkers, and solvers. It stores improvements with the original and corrected actions, checked outcomes, and conditions for reuse. A learned
memory-use policy selects a retrieved record or skips
memory to balance task success and interaction cost, while the base LLM remains fixed. Both
memory and policy are frozen during held-out evaluation. We evaluate COUNTERMEM on 12
benchmark settings across six domains. With gpt-oss-120b, COUNTERMEM improves both ReAct and Reflexion on all 12
benchmarks across six domains, averaging a gain of 12.6 percentage points over their unaugmented versions. In the four-domain comparison across two backbones, task-run
tokens decrease by 7.7-42.0%, excluding offline selector-training costs. Further analyses show that removing verification or persistent storage weakens the gains, while applying verified corrections to unsuitable decisions can reverse them. Code will be released upon acceptance.
Read the full abstract ▾
Computer Vision
CueKFS: Agentic Cue-Driven Keyframe Selection for Long Video Understanding
Weitai Kang, Hanieh Deilamsalehy, Yumo Xu et al. (+3 more)
Keyframe selection (KFS) has long produced compact
video summaries for browsing and
retrieval, and representative frames for thumbnails. More recently, when conditioned on a question, KFS provides an alternative to uniform sampling for long-
video question answering by selecting frames that are more relevant to the question. Most methods rank frames by similarity to the question. Yet a relevant frame may score poorly when the question combines subjects or moments that no single frame shows, or requires implicit information absent from its wording. Other methods try to break down the question into subqueries, but suffer from inaccurate decomposition due to their static initial context. Therefore, we propose CueKFS, a training-free method that reformulates question–frame matching as comparing frames against a set of dynamically generated visual cues. From an initial set of salient frames, we decompose the question into cues. Each cue concurrently probes the
video to navigate to its own evidence. A
reasoning VLM then agentically revises the cue set against its evidence to re-explore the
video. CueKFS then allocates the budget across the surviving cues. Across three
benchmarks, CueKFS establishes state-of-the-art results in all 27 evaluated settings with available prior results, achieving budget-averaged gains of up to +4.54% over the previous baseline and a median of only two VLM calls. We further provide a detailed behavioral analysis of CueKFS, showing that agentic cue refinement drives active re-exploration of the
video, yielding relative similarity gains of up to 92% over the initial context.
Read the full abstract ▾
All authors Weitai Kang, Hanieh Deilamsalehy, Yumo Xu, Dewang Sultania, Serdar Cellat, Yan Yan
Disordered Systems and Neural Networks
Connecting finite-size scaling and renormalization-group flows in Anderson localization on random graphs
Ignacio García-Mata, Weitao Chen, John Martin et al. (+2 more)
We investigate the relation between two complementary descriptions of the Anderson transition on small-world random graphs: finite-size scaling of eigenfunction moments and renormalization-group (RG) flows of multifractal dimensions $D_q$. The RG flows recover the qualitative structure previously reported for the information dimension $D_1$, while their extension to other $D_q$ with smaller moment orders $q<1/2$ confirms the strongly multifractal nature of the localized phase characterized by a clear separation between multifractal and localized behaviors at strong disorder. On the other hand, we show that we can construct another beta function characterizing the flow of eigenfunction moments, whose collapse onto a single function across different disorder strengths provides a clear confirmation of single-parameter scaling property. These results indicate that the family of RG trajectories observed for the information dimension $D_1$ need not imply two-parameter scaling and Kosterlitz-Thouless like critical behavior. To characterize these properties, one should consider different observables, such as large $q>1/2$ and small $q<1/2$ eigenfunction moments, which are associated with distinct critical behaviors, in particular different critical exponents.
Read the full abstract ▾
All authors Ignacio García-Mata, Weitao Chen, John Martin, Bertrand Georgeot, Gabriel Lemarié
Artificial Intelligence
IndustryLLM: Failure-Driven LLM Training for Industrial Procurement
Liang Ding, Zhiang Xu, Yuyang Sheng et al. (+13 more)
Industrial procurement requires
language models to bridge informal buyer jargon, sparse marketplace attributes, and authoritative engineering standards under strict safety tolerances. We present IndustryLLM, an open-weight industrial
language model trained from Qwen3.5-35B-A3B-Base (35B total parameters with ~3B activated per
token, with the
vision encoder frozen). Rather than relying on generic text scaling, we introduce a failure-driven adaptation recipe spanning continued pre-training (CPT) and supervised fine-tuning (SFT). CPT leverages a curated ~100B-
token corpus integrating 5B
tokens of national standards (e.g., GB/T) and technical archives, 10B
tokens of de-identified real-world industrial transaction and inquiry records, and 60B
tokens of general replay. To overcome register mismatch and factual brittleness, we systematically reconstruct an estimated 20B-
token domain subset via multi-register rewriting across 10 genres and 8 writing styles, confidence-routed minimal factual editing, and error-targeted QA synthesis (resolving colloquial typos like '42-luo-mu' -> 42CrMo, expanding ambiguous codes like '16674' -> GB/T 16674, and clarifying conflicting dimensional specs). For downstream deployment, we formalize an evidence-gated constraint-evaluation interface enforcing three-valued logic where unverified product evidence remains unknown rather than satisfied. Offline evaluations demonstrate consistent gains on procurement-query structuring (+2.97 percentage points in exact match, 95% CI [2.11, 3.86] in No-Think mode), while randomized online A/B experiments in production yield substantial improvements (+4.25% GMV, +8.3% satisfied inquiries) alongside a latency reduction from 6-7 s to 1.5 s. Model weights and configs are released at https://huggingface.co/alibaba-
multimodal-industrial-ai/IndustryLLM.
Read the full abstract ▾
All authors Liang Ding, Zhiang Xu, Yuyang Sheng, Bin Chen, Songlin Bai, Run Zhu, Dingjun Wu, Hui Xu, Yandi Wang, Fulin Shi, Leilei Gan, Linlin Yu, Qihuang Zhong, Keqin Peng, Yalong Li, Chengfu Huo
Machine Learning
Deep Reinforcement Learning for Equity Trading: Benchmarking Actor-Critic Methods with Forward Retraining
Bicheng Wang and Xinyi Zhang
Consistently profitable trading is difficult because equity
markets are noisy, non-stationary, and only partially predictable from historical data. We
benchmark five deep reinforcement learning (DRL) actor-critic methods: A2C, PPO, DDPG, TD3, and SAC, that learn trading actions end-to-end from
market states, and compare them with a supervised price-
forecasting baseline. Using daily data for 20 large-capitalization S&P 500 stocks from 2000 to 2020, enriched with trend-following technical indicators and log min-max scaling, we train on 2000-2018 and backtest on 2019-2020. Each
agent is evaluated both when trained once and under forward retraining, in which it is retrained on all data available before each successive test window. DDPG achieves the highest annual return (55.5%), Sharpe ratio (1.38), and alpha (0.22), but also the highest
market beta (1.24). TD3 and SAC offer a better risk-return balance, with Sharpe ratios of 1.37 and 1.33 and maximum drawdowns of about 25%. Forward retraining improves A2C, PPO, and SAC, leaves TD3 essentially unchanged, and reduces DDPG's annual return from 55.5% to 29.8%, consistent with TD3's greater robustness to hyperparameters. The
forecasting baseline has the smallest maximum drawdown (9.6%) and the lowest beta (0.31), underscoring a trade-off between the higher returns of end-to-end DRL and the lower risk of forecast-driven strategies.
Read the full abstract ▾
Sound
CORD-KWS: Calibrated, Order-Aware Detection for Open-Vocabulary Keyword Spotting
Ramesh Gundluru, Adarsh Arigala and Sri Rama Murty Kodukula
Open-vocabulary keyword spotting (KWS) must detect arbitrary keywords without retraining. Cross-
attention-based models achieve state-of-the-art performance but require pairwise interaction between the audio and each keyword, recomputing the fused representation for every audio–keyword pair.
Embedding-based models avoid this computational cost through independent encoding and similarity scoring, yet remain inferior on challenging
benchmarks such as LibriPhrase-hard. We show that this gap is primarily a training issue rather than a limitation of model capacity. While contrastive learning encourages correct keywords to rank above negatives, it does not explicitly calibrate absolute similarity scores, which are critical for applying a fixed threshold across keyword detection. We propose Calibrated, Order-aware Detection KWS (CORD-KWS), a training framework that introduces two complementary objectives while retaining the same encoders, cosine scoring, and O(1) keyword enrollment: a calibrated detection head that supervises absolute scores, and a frame-level CTC objective that preserves phonetic order information. CORD-KWS achieves EERs of 0.43% and 9.64% on LibriPhrase-easy and LibriPhrase-hard, respectively, outperforming existing
embedding- and cross-
attention-based models.
Read the full abstract ▾
Artificial Intelligence
Metro-WM: Long-Horizon Latent Planning with Realisable Sub-Goals
Royson Lee, Fady Rezk, Titouan Parcollet et al. (+2 more)
Model-predictive control with Joint-
Embedding Predictive Architectures (JEPAs) provides a strong zero-shot goal-reaching planner, but it is only effective over short planning horizons. Hierarchical extensions attempt to bridge this gap by learning a macro planner to predict intermediate latent sub-goals to guide the micro planner. In this work, we demonstrate that unconstrained latent sub-goal prediction is fundamentally flawed. A rigorous evaluation reveals that leading hierarchical macro planners routinely emit physically unrealisable sub-goals. To resolve this, we introduce Metro-WM, a hierarchical framework that issues sub-goals by retrieving genuine states from prior experience rather than generating ungrounded latent vectors. Specifically, Metro-WM constructs a graph whose vertices are observed frames from offline expert demonstrations or random-action trajectories, allowing frames from different episodes to be connected and stitched into routes to the goal. Planning over the full graph also makes the system highly robust to execution errors: if the micro planner drifts off course, Metro-WM instantly finds a new optimal path from the current state. Our experiments show that Metro-WM achieves superior long-horizon success rates of up to 37.33 percentage points over the strongest hierarchical baseline on the same world model while being up to 10.9 times faster, requiring both 13-56 times less offline compute and fewer tuned hyperparameters. Additional analysis reveals that Metro-WM frequently finds shorter paths than the offline demonstrations, outperforms an oracle relying on the query's own demonstration on most
benchmarks, and maintains robust performance under extremely sparse
dataset conditions.
Read the full abstract ▾
All authors Royson Lee, Fady Rezk, Titouan Parcollet, Timothy Hospedales, Cristina Cornelio