Peer Revenue Past citations are no guarantee of future returns

Papers

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

All of arXiv

Monday 14 September

Computer Vision

G-ray: Ray-Level Relative Geometric Position Encoding in Multi-View Vision Transformers under Camera Heterogeneity

We study relative position encoding for multi-view vision Transformers under camera heterogeneity, including varying fields of view (FoVs) or projection models. Existing rotary relative position encodings commonly use image-plane positional coordinates, producing projection-dependent relative phases and inconsistent geometric cues for cross-projection attention. We introduce G-ray, a ray-level relative position encoding whose rotary phases are parameterized by camera-local ray angles. The same camera-local ray pair induces the same relative phase across projections, providing projection-invariant positional consistency. G-ray can be used directly or integrated with existing encodings, retaining complementary geometric cues without additional learned parameters. We validate G-ray in three host encodings, RoPE, GTA, and RayRoPE, across 3D reconstruction and novel-view synthesis (NVS). Across three heterogeneous 3D reconstruction benchmarks at 50 views, G-ray leads all six averaged metrics and reduces mean pointmap relative error by 45.8% over MapAnything, with calibration supplied to both. Trained exclusively on homogeneous pinhole images, the 3D reconstruction model handles mixed pinhole and non-pinhole inputs without retraining and remains competitive on homogeneous pinhole 3D reconstruction protocols. For NVS, GTA and RayRoPE improve with G-ray under joint viewpoint and FoV variation. The project's webpage is available at https://g-ray-project.github.io/.
Read the full abstract

All authors Shuo Zhang, Xin Su, Wei Wang, Jun Liu, Xinrui Zeng, Yongsen Chen, Chenjie Wang, Guibo Zhu, Jinqiao Wang, Bin Luo, Liangpei Zhang

Cryptography and Security

PIDS-Bench: Evaluating Prompt-Injection Detectors Under Over-Defense, Obfuscation, and Distribution Shift

Prompt-injection detectors are typically evaluated using aggregate F1 on in-distribution test data, which offers limited insight into behavior under distribution shift, particularly on the benign side of the decision boundary, where false positives impose direct operational cost yet are seldom measured. We present PIDS-Bench, a frozen multi-axis benchmark that jointly evaluates attack detection and benign false-positive behavior at fixed thresholds, spanning in-distribution inputs, hard-benign prompts that mimic injection structure without malicious intent, obfuscated attacks, and domain and structural distribution shifts. We evaluate seven detectors (learned baselines, external prompt-injection classifiers, and broad-safety comparators) alongside a rule-based lower-bound reference. Multi-axis evaluation exposes a failure mode that aggregate F1 conceals. A detector exceeding F1 = 0.98 on the held-out split still misclassifies roughly one-third of an externally-sourced benign subset drawn from public corpora and restricted to security-adjacent content. Across a full threshold sweep and five training seeds, no internal detector reaches an operating point satisfying F1 >= 0.95 and hard-benign FPR <= 0.10 together on this stress distribution. Decomposing by provenance, we find that hard-negative augmentation nearly eliminates over-defense on curated stress inputs but leaves it substantially intact on externally-sourced prompts, a pattern we term provenance-sensitive over-defense. The asymmetry holds across both fine-tuned architectures and does not diminish as the augmentation pool grows, with the externally-sourced FPR remaining far above the 0.10 target. Whether augmentation matched to the externally-sourced distribution would close this gap is untested; threshold calibration and curated-style augmentation alone do not.
Read the full abstract

Combinatorics

Spherical $t$-Designs on $\mathbb S^2$ with $54t^2$ Points

We prove that, for every integer $t\geq 1$, the unit two-sphere admits a spherical $t$-design consisting of exactly $54t^2$ points. More generally, such a design exists with exactly $6q^2$ points for every integer $q\geq 3t$. The proof builds on the topological degree method of Bondarenko, Radchenko and Viazovska, using an explicit equal-area partition of the sphere based on the map of Roşca and Plonka. By choosing the cell centers to minimize the average squared geodesic distance and deriving sharper sampling estimates, we obtain the stated quadratic bound on the size of spherical $t$-designs on $\Sph$.
Read the full abstract

Artificial Intelligence

Four Ledgers, Not One Score: Responsible Communication of LLM-Judge Calibration in Biomedical ML

Synthetic perturbations appear to offer inexpensive calibration data for LLM evaluators in biomedical ML, where expert review is scarce. Yet a planted mutation key is neither a detector output nor automatically human ground truth. We formalize four distinct ledgers: planted perturbations, independent detector outputs, source-linked human dispositions, and human-added discoveries. We then audit the evaluation design, scoring code, read paths, and current human records of a private synthetic Japanese care-handoff workflow. The factory stored 69 planted error cards across 47 targets. Final review covers 22 targets and contains 22 confirmed imported proposals, 9 rejected proposals, and 79 human-added cards; only 3 reviewed targets are double annotated. Passing imported plant keys to a generic detector scorer yields 22/(22+9)=0.710 and 22/(22+79)=0.218. A direct audit identity shows that these values are proposal-confirmation yield and submitted-ledger composition, not judge precision and recall, because no independent detector realization was preserved for the audited proposals in the available records. The audit also finds source-name collisions, row shadowing, forced severity, vacuous ratio defaults, and unsupported zero-support field weights. We contribute a provenance-aware claim audit, a storage contract, and a minimum calibration gate for responsibly communicating biomedical ML capability claims. This single-workflow forensic case is an existence proof of a failure mode, not an estimate of its prevalence: existing human work supports an exploratory audit of synthetic proposals, but not LLM-judge operating characteristics, clinical validity, corpus prevalence, or robust inter-annotator agreement.
Read the full abstract

Robotics

Steering Generative Robot Policies with Lexicographic Preferences

Pretrained generative robot policies can produce effective behaviors across diverse environments, but deployment can lead to requirements and preferences that may not have been represented during training. Furthermore, at deployment, an operator, user, or application may assign these requirements and preferences a priority order that can vary across deployments. For example, embodiment-specific feasibility constraints may need to be satisfied first, while user-specific preferences guide behavior among the feasible options. We show that a frozen generative robot policy—based on either diffusion or flow matching—can be steered at inference time to respect such lexicographically ordered deployment objectives. To achieve this, we introduce two modifications to the sampler. First, we apply dynamic-barrier guidance to sampled trajectories, constraining lower-priority updates so that higher-priority costs do not increase (up to first order). Second, we select the executed sample using a cascade that successively filters candidate samples according to each priority level. The policy weights remain unchanged. On a navigation benchmark, we demonstrate that our method improves success, traversability, and preference compliance over the frozen policy, and achieves substantially better compliance than tuned weighted-sum baselines. The same method transfers to a flow-matching manipulation policy on LIBERO, where it improves compliance without reducing task success. A controlled manipulation study further shows that, in settings where a fixed weight can match the desired ordering, the dynamic barrier reaches comparable best performance over a substantially wider range of parameter settings.
Read the full abstract

Artificial Intelligence

Overflip: Repetition-Induced Label Flips in Guardrail Models

Guardrail models are classifiers deployed to screen malicious prompts and responses in LLM-based services. To meet latency constraints, many lightweight guardrails adopt compact Transformer backbones (e.g., DeBERTa) that are trained with short context windows (typically 512 tokens) and rely on bucketed relative positional encodings to process longer inputs. Prior evaluations assume that a guardrail's decision is stable as the input is lengthened. We show that this assumption can fail. We identify Overflip, a repetition-induced instability where repeating a prompt causes the guardrail's prediction to flip (MAL$\to$BEN) as the sequence grows. We conduct experiments on 9 widely used lightweight guardrail models. Five exhibit MAL$\to$BEN flips on a benchmark of 100 prompts, with confidence margins shrinking steadily with repetition. Among these vulnerable models, flip rates range from 8% to 92%, with first flips occurring at roughly 2.6k–9.4k tokens. Our analysis suggests Overflip differs from traditional attention-dilution baselines, which aim to divert the model's attention away from tokens associated with malicious content, shifting it instead toward unrelated content, such as benign padding or shuffling. While Overflip preserves malicious content, it homogenizes token-level attention over repeated structure and induces a distinct, more gradual attention-dispersion trajectory than padding. Moreover, Overflip poses a greater threat to LLM services than traditional attention dilution methods. Because the bypassed prompt remains semantically intact and is still readily understood by downstream business LLMs, it can transmit malicious intent after passing the guardrail. These findings expose repetition as an attack surface for guardrail models and motivate length-robust evaluation and mitigation.
Read the full abstract

All authors Xu He, Chih-Hsuan Lin, Hung-Mao Chen, Junjie Xiong, Yan Zhai, Kun Sun

Robotics

Atomic Motion Coordinate for Language-Steerable and Force-Responsive Manipulation

Can changing only the language instruction redirect a VLA policy's end effector, or does the visually driven motion prior dominate? We present Atomic Motion Coordinate, a geometry-grounded coordinate for steerable and force-responsive manipulation. Each arm owns thirteen signed translation, rotation, and hold atoms grounded from text and forward kinematics with vision withheld, and the coordinate is injected into every action-expert block via weighted codebook alignment. Contact history modulates the same coordinate through a bounded spherical residual that is recomputed from a fixed nominal latent to regenerate only the unexecuted horizon suffix. Across 7,520 offline horizon interventions, opposite-atom separation reaches 92.5/83.1% (single/dual) versus 39.1/24.0% for LA4VLA-style. Across 50 real-robot trials per task, AMC raises OOD fruit progress from 60.5% to 87.8%; force adaptation raises Plug/Vase from 59.0/71.5% to 78.5/75.2%.
Read the full abstract

All authors Jiaqi Zhai, Jingkai Zhao, Chen Yang, Siyuan Ma, Yutian Zhang, Liwen Yang, Qinglian Wu, Weiqi Fan, Yifei Wang, Yi Zheng, Chenxi Gu, Dong Wei, Wei Zhang

Artificial Intelligence

Semantic-TVM: Structure-Preserving Trustworthy Virtual Memory for Memory-Augmented and Tool-Using Agents

Memory-augmented and tool-using agents expose exact private values when remote LLMs process retrieved memory, tool actions, and intermediate observations. One-way masking limits direct exposure but removes values needed for trusted execution and can leak them through later observations. We propose Trustworthy Virtual Memory (TVM), a closed-loop runtime that keeps exact-value state local while presenting a protected view to the remote model. Within this single runtime, Rule-TVM replaces whole protected fields with locally recoverable handles, and Semantic-TVM instead replaces only sensitive spans predicted by a trusted local model, preserving surrounding task-relevant context. On Memory-EHR and Memory-RAP across two providers, span-level projection recovers most of the EHR utility lost under whole-field replacement (Task Success 84.17% vs. 52.33% on DeepSeek) while measured exposure stays low and workflows remain executable.
Read the full abstract

All authors Yu Li, Qikun Cai, Tao Huang, Chen Hou

Quantitative Methods

Telegraph Processes with Extrinsic Fluctuations: Burst Dynamics and an Application to Domestic Cat Activity

Animal activity often consists of intermittent bursts separated by prolonged periods of inactivity. Here, we describe this behavior using a two-state telegraph process with extrinsically fluctuating transition rates. We derived analytical expressions for the stationary behavior of the system and characterized how stationary fluctuations in the effective transition rates modified the activity and residence time statistics. In particular, we show that whereas fixed transition rates lead to exponential residence-time distributions, stationary rate fluctuations generate effective Lomax distributions with heavier tails. We also found that fluctuations can either increase or decrease the mean occupancy of the active state, and that when the fluctuations in the two transition rates are equal, their effect on the stationary occupancy becomes indistinguishable from the case without fluctuations. We assessed the parameter inference using synthetic data and applied the framework to observations of a domestic cat. For synthetic data, accounting for extrinsic fluctuations substantially improved the recovery of the kinetic parameters used to generate the data. In the experimental data, the inferred variability of the inactive-to-active transition rate was substantially larger than that of the active-to-inactive rate. These results establish a tractable framework for studying stochastic switching systems in which extrinsic variability reshapes burst and residence-time statistics.
Read the full abstract

Artificial Intelligence

CoMem: Collective-Individual Memory Synergy for Evolutionary Multi-Agent Systems

Designing effective memory mechanisms is crucial for advancing LLM-driven Multi-Agent Systems (MAS), helping agents learn together and perform better over time. While recent work has led to strong cooperation skills, most methods still use flat, unstructured memories, which easily get filled with noise and erase differences between agents. To address this, we introduce the concept of collective-individual memory synergy and propose CoMem, an architecture that unifies both private experience and shared knowledge for multi-agent learning. CoMem features:(i) Private Experience Sedimentation, which lets each agent keep and update its own useful memories over time;(ii) Collective Wisdom Curation, which carefully selects only widely proven ideas to be shared among agents;(iii)Parallel Dual-Stream Retrieval, which allows agents to draw both from their own memory and the group's wisdom, using clustering to ensure diversity.Experiments on ALFWorld and PDDL benchmarks show that CoMem achieves strong overall performance and robustly avoids memory pollution.
Read the full abstract

All authors Chengxin Yu, Zhaoxin Fan, Faguo Wu, Hongwei Zheng, Yun Zhou, Zhiyu Li

Logic

Higher-Order Quadripolar Argumentation Framework and Encoded Semantics

This paper introduces the Higher-Order Quadripolar Argumentation Framework (HQAF), which extends bipolar frameworks by allowing attacks and three types of supports—necessary, deductive, and evidential—to interact at higher order. In HQAF, each interaction may have as its source and target an argument, an attack, or a support. We provide a 3-valued semantics via adjacent complete labellings and equational systems, and define encoded semantics directly in propositional logics, mainly Łukasiewicz three-valued logic for the 3-valued case and fuzzy logics based on continuous t-norms (Gödel, Product, Łukasiewicz) for the fuzzy case. Model equivalence between HQAF and its normal encoding is established. The fuzzy equational semantics is shown to be a specific formalism of the fuzzy encoded semantics, and ternarization connects fuzzy models to 3-valued complete labellings. The study is confined to non-set HQAF; collective interactions are reserved for future work.
Read the full abstract

Computation and Language

ABSOL: Aggregated Bayesian Subsampling Orchestrated with LLMs

Large language models are increasingly used as natural-language interfaces to structured data, yet they remain unreliable when answers require consistent evidence conditioning, dependency-aware reasoning, and uncertainty estimation. Bayesian networks provide an explicit probabilistic reasoning layer, but learning useful structures from data remains costly and fragile at scale. We introduce ABSOL, a hybrid LLM-guided Bayesian network structure-learning framework that uses LLMs as bounded semantic guides. Across five discrete BN benchmarks spanning 27 to 1041 nodes, ABSOL is the only evaluated method to produce a viable graph on every benchmark, and achieves the highest Edge F_1 on every benchmark larger than 27 nodes with GPT-5.4. The four LLM augmentations, which contribute complementary semantic evidence to the statistical backbone, improve Edge F_1 over the non-LLM aggregation backbone by +0.23 on average. Complementary post-hoc refinement experiments suggest that these gains depend in part on limiting the LLM's authority over the final structure. Together, these results show that language-derived semantic knowledge can substantially improve scalable probabilistic structure learning when used as bounded guidance within a statistically grounded reasoning pipeline. The code for ABSOL is available at github.com/megagonlabs/absol-bn.
Read the full abstract

Nuclear Theory

Inclusive Neutral-Kaon Photoproduction on the Deuteron

We report on our study of inclusive neutral-kaon photoproduction on the deuteron, $d(\gamma,K^0)YN$, for photon energies between 0.9 and 1.1 GeV. The calculation is performed in the impulse approximation, with the deuteron wave function generated from the Bonn One-Boson-Exchange-Potential in $q$-space (OBEPQ) within a non-relativistic framework, while the kinematics and the elementary operator are kept relativistic. For the elementary operator we employ a recently developed isobar model that includes high-spin nucleon and $\Delta$ resonances and has been constrained by nearly 20,000 data points, and we compare its predictions systematically with those of the Kaon-Maid model. For the first time, the NKS1 and NKS2 data are subjected to a quantitative analysis within a modern elementary production framework. The New Operator provides a considerably better description of the data, whereas Kaon-Maid overestimates the measured cross section by up to a factor of four at the higher photon energies, a discrepancy that can be traced back to its unconstrained $\gamma n \to K^0\Lambda$ amplitude. We further show that the finite photon-energy and kaon-angle bins of the existing measurements generate a theoretical uncertainty comparable to the difference between the two models, so that a proper comparison with the data requires the calculation to be averaged over the experimental acceptance. A three-dimensional mapping of the cross section over the kaon momentum and angle reveals a narrow quasi-free ridge, accompanied by a second structure associated with the opening of the $\Sigma$ channels, from which we identify the kinematics most favorable for future measurements. Finally, the tensor target asymmetries are found to be far less sensitive to the elementary operator than the cross section, and therefore probe the nuclear dynamics in a way that is complementary to the cross section.
Read the full abstract

Robotics

IMPACT-VLA: Interaction-aware Multimodal Propagation Attribution via Counterfactual Trajectories for Vision-Language-Action Policies

Vision-Language-Action (VLA) policies perform robot manipulation tasks using multimodal inputs such as visual observations, proprioceptive states, and language instructions. However, it remains unclear at which execution stages each modality contributes to final task success and how input interventions propagate through subsequent states, observations, and actions. Existing attribution approaches primarily measure local sensitivity or temporally aggregated importance, limiting their ability to capture phase-dependent contributions and cross-phase dependencies. We propose Interaction-aware Multimodal Propagation Attribution via Counterfactual Trajectories for Vision-Language-Action Policies (IMPACT-VLA). IMPACT-VLA constructs behavioral phases from action transitions in a successful reference rollout, aligns them with policy query boundaries, and defines phase-modality blocks as attribution units. It then performs closed-loop counterfactual re-execution to quantify each block's contribution to final task success. We further analyze cross-phase non-additive interactions and trajectory propagation while distinguishing behavioral from functional recovery. Across 30 LIBERO robot manipulation tasks using OpenVLA-OFT, dominant-modality transitions occurred in 25 tasks (83.3%), and closed-loop attribution identified task-critical information more faithfully than Static Action Perturbation. Later-block marginal gains for negatively interacting pairs increased by approximately 3.3x under early-phase input replacement, while functional recovery could occur without behavioral recovery. These results reveal when multimodal inputs support task success and how their contributions become conditionally coupled during closed-loop execution.
Read the full abstract

Quantum Physics

Failure of the Asymptotic Equipartition Property for Quantum Channels

We establish two results on the asymptotic equipartition property (AEP) for quantum channels. First, the AEP formulated using smoothing over completely positive trace-preserving maps fails in general. Second, the AEP with smoothing over subchannels (completely positive trace-nonincreasing maps) is equivalent to the strong converse for parallel quantum channel Stein's lemma, to a sharp threshold for the channel hockey-stick divergence, and to an AEP for the channel Lorenz divergence. The counterexample reveals a surprising separation between channel and subchannel smoothing. Although states and classical channels admit a sharp, dimension-independent completion bound, which we establish, the corresponding single-shot gap for quantum channels is unbounded already for qubits. This separation persists asymptotically: we construct qutrit channel pairs with finite max-relative entropy whose CPTP-smoothed rate strictly exceeds the regularized channel relative entropy. The asymptotic gap between channel and subchannel smoothing is unbounded across the family, even when the subchannel approximation error decays exponentially. To establish the equivalences, we develop a uniform filter that converts Lorenz smoothing into a single nearby subchannel. The four conjectured formulations thus share a common asymptotic threshold, whose validity for general channel pairs remains open. Together, these results separate this threshold question from the obstruction imposed by exact trace preservation.
Read the full abstract

Logic

The Borel Distinguishing Number of Schreier Graphs

The Borel distinguishing number $D_B(\mathcal{G})$ of a Borel graph $\mathcal{G}$, recently introduced by Bilge and Kaya, is the minimum number of colors required to break the symmetry of $\mathcal{G}$ in a Borel way. In this paper, we investigate the Borel distinguishing number of Schreier graphs induced by the free part of the shift action $\Gamma \curvearrowright n^\Gamma$. We prove that $D_B(\mathcal{G})\le n+1$ for $\Gamma=\mathbb{Z}^d$ equipped with the standard generators. Moreover, we show that $D_B(\mathcal{G})\ge n+1$ if $ \Gamma$ is amenable and $ \{\gamma \in \mathrm{Aut}(\mathrm{Cay}(\Gamma,S)) \mid \gamma(e) = e \}$ is non-trivial. We also show that $D_B(\mathcal{G})$ is finite if $\Gamma$ is finitely generated, and give some applications of our results. These results answer some questions raised by Bilge and Kaya.
Read the full abstract

Astrophysics of Galaxies

Nexus-CDM: Isolated Galaxy Simulations with Cosmologically Evolving Dark-Matter Halos I. Method and Validation

Galaxy evolution can be modelled in two complementary ways. Standalone simulations place a single galaxy in isolation, evolving it under controlled initial conditions without the surrounding cosmic web. This approach offers high resolution and computational efficiency, making it well suited to disentangle specific physical processes (e.g. feedback and disc instabilities) and testing them cleanly, though at the cost of ignoring environmental context, gas accretion from the broader large-scale structure, and the hierarchical assembly that feeds real galaxies over cosmic time. Cosmological simulations, by contrast, form galaxies self-consistently within an expanding universe, capturing large-scale structure formation, mergers, gas inflows, and environmental effects such as tidal stripping and ram-pressure. This realism comes at a steep computational cost, requiring coarser resolution or simplified sub-grid physics for star formation and feedback. Here, we explore a new paradigm that combines the relative merits of both methods – what we call the Nexus-CDM framework. This approach naturally incorporates galaxy mass growth, which is absent from existing standalone simulations, reinforcing the limitations of static halo models used in galaxy formation and evolution research. The viability of the framework is demonstrated in an idealised setup, where it successfully reproduces the formation and subsequent evolution of a realistic stellar disc. We present some early results, in particular, discs form readily in a gravitational potential with a shallow central gradient, contrary to recent claims.
Read the full abstract

All authors Thor Tepper-García, Joss Bland-Hawthorn, Oscar Agertz, Christoph Federrath

Machine Learning

HGTO: A Unified Graph-Based Physics-Informed Formulation for Structural Topology Optimization

Density-based topology optimization is typically structured as a nested sequence of material updates, structural analyses, and sensitivity assessments. While neural density parameterization and dual-field physics-informed approaches provide data-free alternatives, most existing methods represent density and displacement as coordinate fields and make limited use of the discrete relationships inherent in the finite element mesh. The present study introduces HGTO, a unified graph-based formulation that extends complete neural topology optimization from coordinate space to finite-element graph space. Element densities are parameterized on the element graph derived from the mesh, and the structural state is determined on the corresponding node–element hypergraph. Finite element kinematics, numerical quadrature, constitutive response, and force assembly remain explicitly defined operations within the differentiable computation. The material field and equilibrium state are therefore coupled through a common finite-element incidence structure. Numerical studies show compliance comparable to conventional density-based optimization at substantially lower computational cost than a representative coordinate-based dual-field neural method. The same coupled formulation accommodates high-resolution and irregular meshes, three-dimensional structures, finite deformation, and elastoplastic response.
Read the full abstract

Human-Computer Interaction

Exploring Avatar-Based Representations of Desktop Analytical Workflows for Asymmetric Collaborative Visual Analytics

Collaborative visual analytics increasingly occurs across asymmetric desktop-VR settings, with desktop analysis offering precision and efficiency and VR providing spatial and embodied affordances. However, maintaining workspace awareness remains challenging because desktop collaborators are often represented in VR only through indirect cues such as perspective sharing, shared visualization state, or lightweight cursor traces, which do not convey their ongoing analytical activity in a VR-native manner. To address this gap, we present Desk2Avatar, which explores avatar-based representations of desktop analytical workflows in VR. We introduce two representation strategies, Distanced and Embodied, inspired by remote pointing and direct manipulation. We conducted a within-subject study with 18 participants comparing these strategies against a depth-adaptive cursor baseline. Our findings show that avatar-based representations improved collaborator awareness and attentional guidance over cursor cues, while excessive embodiment introduced occlusion, distraction, and additional workload. Finally, we discuss design implications for future desktop-to-VR representations, focusing on balancing collaborator presence, attentional guidance, and workspace readability.
Read the full abstract

High Energy Astrophysical Phenomena

Optical spectroscopy of the VHE flat-spectrum radio quasar PKS 0903-57

We report new optical spectroscopic results for the gamma-ray blazar PKS~0903$-$57. The optical properties of this source have historically been complicated to establish, due to a bright foreground star separated from the blazar by only 0.67$^{\prime\prime}$. During its strong high-energy (HE) and very high-energy (VHE) gamma-ray flaring period between 2020 and 2022, the Southern African Large Telescope and Very Large Telescope (VLT) took spectra of its optical counterpart. Photometric monitoring with the Rapid Eye Mount (REM) telescope complemented the observations. The observational results were inconclusive. A major improvement was achieved with a repeat of the VLT observations in 2024 using a 0.5$^{\prime\prime}$ slit under excellent seeing conditions. This led to the detection of five narrow emission features in the spectra, yielding a robust redshift measurement of $z$ = 0.2621 $\pm$ 0.0006. We also detected a broad symmetric H$\alpha$ emission component with a full width at half maximum of 4020 $\pm$ 30 km/s in the spectra. Non-thermal jet emission was found to dominate the continuum, supporting the classification of PKS~0903$-$57 as a flat-spectrum radio quasar. Another interesting observation was a significant velocity displacement of $\sim$1500~km/s between the broad H$\alpha$ line and its corresponding narrow line. Such a displacement may be associated with unusual dynamics in the broad-line region or accretion flow, a system containing two supermassive black holes, or a recoiling black hole produced following a supermassive black-hole merger. Building on these results, we carried out additional VLT, REM, and Swift observations towards the end of 2024 and during 2025. We present in this paper the preliminary results from the analysis of these new data.
Read the full abstract

All authors E. Kasai, P. Goldoni, C. Boisson, S. Pita, W. Max-Moerbeck, F. D'Ammando, B. Rajput, M. Backes

Artificial Intelligence

Shallow Beliefs: Synthetic document finetuning does not inoculate against emergent misalignment from reward hacking

Recent work shows that models that learn to reward hack on RL environments can become broadly misaligned, and that reframing reward hacking as acceptable behavior during training (inoculation prompting, or IP) blocks this generalization. We ask whether synthetic document finetuning (SDF) can inoculate a model against future training we don't intervene on. We add synthetic documents framing reward hacking as acceptable behavior to a model's midtraining corpus, and then train these models with RL on exploitable environments, teaching them to reward hack. Behaviorally, midtraining succeeds: models describe reward hacking favorably and are more approving of reward-hacking outputs they produce. However, they show strong EM after learning to reward hack, while IP in the same setting prevents EM. We show that SDF can predictably steer downstream generalization when inserting new associations, but struggles and has unpredictable effects when overriding existing associations, such as that between reward hacking and misalignment that produces EM. Our results suggest that, at the scales we test, SDF can make a model appear aligned with desired beliefs while steering its generalization from later training in unintended ways.
Read the full abstract

Robotics

From Learned-Mode AV-Traffic Pairing to Planner Decisions: A Marginal-Preserving Study on Argoverse 2

Joint motion forecasts pair each autonomous-vehicle (AV) future with surrounding traffic, but actor-level metrics do not show whether that structure matters to a planner. We study this question with a marginal-preserving product control that removes AV-traffic pairing among the learned modes while retaining the fixed constant-velocity pair and holding trajectories, actor-level marginals before planner conditioning, candidates, the cost terms and weights, and fallback fixed. The intervention also changes candidate-conditioned concentration. Across twelve runs on 1,400 held-out Argoverse 2 scenarios, the intervention changes 3.0% of route-level offline selections at $\tau=4$ m. Control-minus-joint recorded-trajectory regret is $-0.026$ and $-0.118$ at the two training sizes; crossed and seed-$t$ intervals span zero. At $\tau=1$ m, relative costs change in 87.9% of route evaluations and route-level offline selections in 8.1%. Before concentration matching, descriptive outcome estimates favor the control. Most of this gap disappears along an approximate concentration-matching path; the remaining contrasts are $+0.112$ and $-0.047$, and both crossed intervals span zero. Actor-level forecast metrics remain identical. Pairing-strength and temperature sweeps show that the decision contrast grows with pairing removal and sharper conditioning. The intervention changes planner decisions even though actor-level metrics remain unchanged. The matching analysis, however, cannot separate any recorded-outcome effect of learned-mode AV-traffic pairing from the accompanying change in conditioned concentration.
Read the full abstract

Exactly Solvable and Integrable Systems

Multivariable Painleve'-II equation: connection formulas for arbitrary system size

Connection formulas for the asymptotic solutions of a system of n> 1 coupled Painleve'-II equations with symmetry-breaking parameters are written explicitly. An asymptotically exact WKB approach to these formulas relies on the quantum-mechanical independent crossing approximation for an explicitly time-dependent Schroedinger equation.
Read the full abstract

Computation and Language

Intelligence Under Time Constraints: Rethinking Test-Time Compute

Intelligence under time constraints requires deciding not only how much to compute, but when computation is worth starting. We study this problem in streaming interactions, where evidence arrives incrementally and may be revised. Early computation has more time to finish but rests on incomplete evidence; waiting improves information while shrinking computational slack. We call this the information-slack dilemma. We take the evidence-dependent computational job as the unit of analysis: when to start it, what supports its result, and when that result can be committed. Advance computation is valuable only insofar as its benefits survive the costs of verification, invalidation, and recovery. This applies to grounded incremental processing and reusable preparation as well as future-dependent speculation. We propose a research agenda on computation under evolving evidence, prioritizing selective recovery under controlled evidence revisions. Evaluation should separate earlier-execution effects, deployment value against a full-input alternative, and the added value of predictive policies, while accounting for shared-resource costs. The objective is not maximal advance computation, but more trustworthy, on-time responses within a declared resource envelope.
Read the full abstract

Strongly Correlated Electrons

Heisenberg-like critical behavior in the quasi-two-dimensional metallic ferromagnet LaCrSb3

LaCrSb3 is a material exhibiting both quasi-two-dimensional spin fluctuations and three-dimensional magnetic interaction characteristics. By measuring the isothermal magnetization of single-crystals and conducting a systematic critical behavior analysis, we clarify the critical properties of its ferromagnetic phase transition and the intrinsic magnetic interaction mechanism. Based on high-precision isothermal magnetization data measured in the vicinity of the critical point, the Curie temperature for the ferromagnetic-paramagnetic phase transition is determined to be TC = 126 K, with the critical exponents obtained as $\beta$ = 0.376, $\gamma$ = 1.417 and $\delta$ = 4.76 via the self-consistent iterative method based on the Arrott-Noakes equation. The reliability of these critical exponents is verified by the Widom scaling law, the magnetic state scaling equation and other analyses. A comparison with theoretical models demonstrates that the critical behavior of the magnetic phase transition in this system basically belongs to the universality class of the three-dimensional Heisenberg model. This conclusion is further confirmed by the distance-dependent decay behavior of the exchange interaction J(r), revealing the dominant role of isotropic direct exchange interactions in this system. Finally, drawing on research findings of other quasi-two-dimensional magnetic materials, this work proposes that LaCrSb3 may exhibit finite-temperature magnetic order in the two-dimensional limit, thereby possessing important theoretical research significance and promising practical application prospects.
Read the full abstract

All authors Qianhui Mao, Haotian Deng, Bin Chen, Jinhu Yang

Mesoscale and Nanoscale Physics

1D Interface-Induced Real-Space Berry Curvature Gradient Encoded in SpinValley-Coupled Nonlinear Photocurrent in Lateral Heterostructure

Symmetry breaking enables helicity-dependent optoelectronic responses in quantum materials. In two-dimensional transition-metal dichalcogenides (TMDs), circular photocurrent (CPC) is typically observed under oblique illumination via the circular photon drag effect, while realization of the intrinsic circular photogalvanic effect (CPGE) under normal incidence remains challenging. Here, we report a clear signature of CPGE in chemical vapor deposition-grown monolayer 2D lateral heterostructure (LHS). By probing the nonlocal photocurrent across the 1D hetero-interface, we observe a helicity-dependent photocurrent consistent with a spin-valley-coupled CPGE-dominated mechanism driven by interface-induced Berry-curvature asymmetry, i.e., Berry curvature dipole. Spatially and spectrally resolved measurements under selective excitation of the MoSe2, WSe2, and the interface regions exhibit distinct magnitudes and polarities of the CPG responses. These contrasting behaviors highlight the opposite spin-valley dichroism for MoSe2 and WSe2, as well as enhanced valley mixing at the 1D interface. Our findings establish LHS as a robust platform for realizing a hetero-interface-induced CPGE-like response and highlight its potential for valleytronics and optospintronics.
Read the full abstract

All authors Suman Kumar Chakraborty, Jinu James, Biswajeet Nayak, David Tebbe, Purbasha Ray, Christoph Stampfer, Bernd Beschoten, Lutz Waldecker, Prasana Kumar Sahoo

Computation and Language

MTAC-IFBench: Benchmarking Instruction-Following in Multi-Turn Agentic Coding

Recently, the rapid development of large language models (LLMs) has reshaped software engineering by enabling autonomous code agents that plan, execute, and utilize external tools iteratively to tackle complex tasks. Beyond achieving functional correctness, these agents must faithfully follow process instructions and constraints throughout the development lifecycle. However, existing benchmarks typically focus on final functional correctness or confine instruction-following evaluation to single-turn, general chat or simple code generation scenarios, leaving instruction-following in multi-turn agentic coding underexplored. To bridge this gap, we propose MTAC-IFBench, a comprehensive benchmark for this critical capability. It features multi-turn progressive software development instructions with diverse constraints spanning 6 primary and 18 secondary categories. With an average of 7.04 turns and 91.33 constraints per instance, it poses a rigorous challenge to current LLMs. To make the evaluation reliable, we construct a checklist for each constraint and functional requirement, and integrate verification scripts and judge agents to verify each checklist item. MTAC-IFBench identifies significant deficiencies in existing code agents in multi-turn instruction-following, with their performance degrading rapidly as the interaction session grows longer.
Read the full abstract

All authors Bosi Wen, Cunxiang Wang, Jiayi Gui, Haoke Zhang, Yilin Niu, Pei Ke, Dayong Yang, Hongning Wang, Minlie Huang

Computation and Language

Typhoon ASR Streaming: Steerable Low-Latency Thai Speech Recognition with Real-Time Shallow Fusion

Open Thai automatic speech recognition (ASR) is dominated by offline, Whisper-based models that read the whole utterance before transcribing, ruling out low-latency uses such as live captioning and voice agents. We present a deployable system for streaming Thai ASR that lets a user steer its vocabulary at decode time, without retraining. A widely used open Thai model, trained with full context, collapses when run as a true stream; we restore streaming with a cache-aware encoder, by converting it or adapting a natively streaming one, and add a shallow-fusion layer that re-ranks candidates inside the streaming decoder with a GPU n-gram language model and phrase boosting. Across two Thai benchmarks and two model sizes, the streaming models stay usable where the full-context model fails, cutting character error rate 4.3-4.5x at a one-second look-ahead while running faster than real time. Decode-time steering then lifts keyword recall from 16.6% to 20.7% at no accuracy cost and negligible overhead; most of the gain comes from an n-gram over ordinary training transcripts, which resolves the written form of code-switched words the model hears but spells inconsistently, with phrase boosting adding targeted control over rare domain terms.
Read the full abstract

All authors Warit Sirichotedumrong, Tanawin Samutsin, Shah Faisal Wani, Sittipong Sripaisarnmongkol, Kunat Pipatanakul

Audio and Speech Processing

Learned Bow Control on a Measured Bowed-String Model: a Revised Minimum-Bow-Force Law, a Recurrent Controller, and the Domain of a Supervision Ceiling

A finite-difference bowed-string model with implicitly resolved Stribeck friction is presented, with a regime diagnostic. Without implicit resolution no stick phase forms at any bow force. With friction, impedance and quality factor from published measurement rather than fitted, all four strings return a stick fraction of 89.1% against an ideal 90%. Schelleng's maximum bow force is recovered on every string. His minimum, $F_{\min} \propto Z^2 v_b \beta^{-2}$, is replaced by a law, $F_{\min} = C Z v_b / \beta$ with a dimensionless $C = 1.112 \pm 0.017$ that reproduces on sixty-four held-out operating points. Six controllers at matched capacity, on four strings at twenty seeds, place a gated recurrent network ahead of a feedforward one, its margin over the mean of the other five largest when the commanded bow speed is overridden at mid-stroke. The feedforward network completes more strokes, mostly from a cold start no player would use. A minimal gated variant fails because gates computed from the input alone cannot clear a latched state. Training loss selects neither the capacity nor the context length. No learned controller improves on the lookup rule that generated its labels where that rule is correct. Of the rule's 1089 cells, 57 are playable on a properly settled plant and unplayable by its labels, and the controller commands them where the lookup will not. The controller's score regresses on the rule's with a slope of 0.32, more than ten standard errors below unity, so it overtakes the rule where the rule fails and is bounded by it where it holds. Under a rigid finger stop the plant is provably invariant, so transfer loss between pitches is the controller's alone, traced to one feature. A regime classifier without a stick test mistakes small-amplitude periodic slipping for Helmholtz motion, and a harmonicity measure rates a string the bow never grips above it.
Read the full abstract

Quantum Physics

Correlation and Branching Mechanisms in the Quantum-to-Classical Transition of Interacting Systems

We analyze the quantum-to-classical transition of interacting systems from the viewpoint of reduced dynamical closure. Bohmian trajectories are used as a diagnostic of the probability flow generated by the full wavefunction, rather than as an alternative set of quantum predictions. A closed classical or semiclassical description of a retained coordinate can emerge either because interaction-induced correlations are suppressed or because the observable of interest is insensitive to the correlations that remain. Three models are used to isolate distinct mechanisms. In a hard-core collision, the leading tail of a broad relative-coordinate wave packet reaches the boundary before its center and generates spatial nonseparability through reflection and interference. In the Rabi model, the mixed quantum-classical Ehrenfest approximation fails when spin-position covariances prevent the spin dynamics from being determined solely by the mean oscillator coordinate. In an internal-state-dependent force model, the total mean position obeys an exact Ehrenfest equation, while the wavefunction separates into branch-resolved pointer components; mean classicality therefore does not imply branch-level classicality. These examples show that the quantum-to-classical transition of interacting systems is controlled not only by wave-packet localization but also by whether and how interaction-induced correlations enter the reduced equations of motion.
Read the full abstract