Papers

We investigate the classification of $C^*$-algebras generated by pairs of isometries $(V_1, V_2)$ satisfying $V_1V_2=qV_2V_1$, where $q=e^{2\pi i\theta}$ with $\theta$ irrational, up to unitary equivalence. Using a functional model that realizes pure $q$-commuting pairs on vector-valued Hardy spaces in terms of a uniquely associated triple $(\mathcal{F}, P, U)$ consisting of Hilbert space $\mathcal{F}$, an orthogonal projection $P$ and a unitary operator $U$, we analyze the internal structure of $C^*(V_1,V_2)$ via an extension of the $C^*$-algebra generated by its minimal $q$-commuting unitary extension by the defect ideal $\mathcal{I}_{\mathcal{D}} =\langle [V_1^*, V_1],[V_2^*, V_2]\rangle$. For pairs that are essentially doubly $q$-commuting, we show that $PUP|_{\operatorname{ran} P}$ is semi-Fredholm and prove that $|\operatorname{ind} PUP|_{\operatorname{ran} P}|$ is invariant under unitary equivalence of the corresponding $C^*$-algebras, thereby providing an explicit criterion for distinguishing non-equivalent pairs. This work extends the seminal 1978 results of Berger, Coburn, and Lebow for commuting isometries, while addressing the key technical nuances of this noncommutative setting: most notably, navigating the failure of Gelfand theory due to noncommutativity and replacing the classical Toeplitz extension by an extension of the noncommutative torus by the ideal of compact operators.

Background. Agentic AI systems independently decompose tasks such as literature search, data analysis, and programming into subtasks, search the web, access databases, and execute code. This allows them to perform digital research tasks at high speed. Objectives. Under what conditions does the use of agentic systems produce reliable efficiency gains, and which tasks remain with researchers? Materials and methods. Summary of current studies on literature searches, data analysis, software development, and clinical decision support. Results. For digital activities, work shifts from execution to steering and review. Efficiency gains are greatest when expected behavior can be formalized in advance and tested automatically. In complex agentic systems, recorded sequences of reasoning steps and tool calls can quickly become too extensive for human review. Furthermore, explanations generated by the model do not reliably reflect how an output was produced. One possible step toward more reliable systems is the validation of individual components. The limited reviewability extends beyond research itself; the review of scientific articles and grant proposals is also reaching capacity limits. In pathology, curated and annotated data, researchers' own analytical skills, and institutional exchange of experience are becoming increasingly important. Conclusions. Researchers remain responsible for their results. They must determine what to delegate and how to review the results. The importance of a research question to patients, the field, and society cannot be fully assessed using formalized criteria and remains a matter of expert judgment. Agentic systems can free up time for this.

This article develops a physics-informed deep learning framework for single-snapshot 3D phase-only positioning in narrowband phase-coherent distributed multiple-input multiple-output (D-MIMO) networks under two-ray propagation. Unlike prior phase-coherent D-MIMO localization methods that largely assume line-of-sight (LoS)-only channels, we consider a LoS path and a specular ground reflection. Since the narrowband single-snapshot observation cannot resolve these components in delay, their coherent superposition induces structured perturbations in the carrier phase measurements. To address this, we develop a Gaussian process (GP)-based model that learns the quasi-periodic phase distortion caused by ground reflection from carrier phase measurements collected at a small number of training locations and uses it to generate high-quality synthetic samples. Building on these samples, we propose the Phase-Only Positioning Transformer (POPT), an encoder-only transformer that captures inter-antenna point (AP) phase relationships without solving the highly non-convex maximum-likelihood (ML) problem underlying model-based estimators. We also derive the fundamental position error bound (PEB) and develop maximum-likelihood estimation (MLE) for the considered multipath phase-only D-MIMO positioning problem. Numerical results show that, with only 50 GP training locations, the proposed method achieves near-PEB accuracy. The analytical MLE is highly sensitive to relative permittivity mismatch, whereas the proposed method mitigates this dependence by learning the phase perturbation directly from measurement data. Compared with MLE, the proposed approach also reduces floating-point operation (FLOP) complexity and inference time by about 1.7 and 3.6 orders of magnitude, respectively.

Misbehavior detection in vehicle-to-everything (V2X) systems is essential for ensuring the semantic correctness of exchanged messages and preventing the dissemination of falsified information. Existing data-centric misbehavior detection approaches largely rely on statistical validation or supervised machine learning models under the implicit assumption of fully observable V2X streams. In practice, however, vehicular environments are inherently partially observable due to hardware failures, intermittent connectivity, and environmental occlusions. Moreover, missingness itself can be strategically exploited by adversaries to evade detection. In this paper, we study misbehavior detection under incomplete V2X observations and propose a deep reinforcement learning (DRL)-based detection framework that learns adaptive policies with incomplete data. We further introduce an adversarial threat model in which attackers exploit or deliberately induce missingness to evade detection, including evasion via natural occlusions and adversarial feature suppression. Extensive experiments conducted on the VeReMi dataset under various missingness patterns demonstrate that DRL significantly outperforms a powerful XGBoost baseline under natural partial observability. However, results also reveal a critical vulnerability: DRL policies can be highly susceptible to evasion attacks that strategically exploit natural missingness. In contrast, DRL exhibits more gradual degradation under direct feature suppression compared to static tree-based models.

We prove Koebe's conjecture for John domains by showing that every planar John domain is conformally equivalent to a circle domain. Our proof is based on a geometric property of John domains, which we prove to be sufficient for uniformization by a circle domain. As a consequence, we obtain an alternative proof of Koebe's conjecture for inner uniform and Gromov hyperbolic domains, recovering results established in earlier work of the authors. Finally, we derive a necessary and sufficient condition for a domain to satisfy Koebe's conjecture by using Schramm's transboundary modulus and well-behaved exhaustions of circle domains constructed by the second author and Rajala.

Large language models (LLMs) as a judge enable scalable evaluation, but their judgments can be sensitive to response order and, even after removing such position effects, can still diverge systematically from human preferences.We introduce DIAL, a unified framework that combines abundant LLM comparisons with limited human comparisons to separate judge-specific position effects, learn shared structure in position-debiased LLM preferences, and adaptively calibrate that structure toward the human preference target. Theoretically, we study three aspects of DIAL: (i) identification of latent LLM preferences, position effects, and human calibration; (ii) adaptive estimation that balances LLM anchoring against limited human evidence; and (iii) fixed-weight uncertainty quantification for the calibrated human preference. Empirically, we evaluate position debiasing and human alignment separately in controlled simulations and on three human-preference benchmarks, showing that DIAL remains robust to unbalanced response order, achieves strong human-aligned rankings with limited labels, and adapts toward human evidence when LLM information is imperfect. Our real-data study collects over 410K judgments from 21 LLM judges in both display orders, providing a resource for future studies of LLM-judge bias, heterogeneity, and human alignment.

Estimating the influence of training examples on model behavior is essential for data debugging, valuation, and attribution. Existing influence estimators often produce incompatible rankings, which are commonly ascribed to approximation error. We argue that a more fundamental source of disagreement is specification mismatch: influence depends on the behavior being attributed, the intervention applied to each training example, and the counterfactual training process that maps the intervention to a model response. These choices are especially important when the target behavior requires a tractable surrogate, such as query loss, a logit, or a margin. We formalize influence as a counterfactual estimand, distinguish specification mismatch across estimands from approximation error in estimating a fixed estimand, and organize representative estimators by their implied specifications. We further derive a local decomposition that exposes how behavior signals, training signals, and counterfactual parameter responses interact. Controlled experiments show that exact estimands under different specifications can induce different rankings, whereas approximation error grows as perturbations move farther from their linearization points. Experiments on noisy label detection and LLM attribution show that specification choices significantly affect attribution quality, especially for the choice of behavior surrogate. Behavior-aligned specifications can identify target-specific training examples obscured by default loss-based or similarity-based specifications. These results establish specification analysis as a necessary first step for interpreting and comparing data influence estimators.

A ball bouncing on a thicker cushion should bounce less. We show that this is not always so . Measuring the coefficient of restitution $e$ of a steel ball dropped on a stack of $N$ sheets of standard A4 printer paper, we find that $e$ first drops, then rises to a pronounced maximum, and only then decreases towards a minimum. The maximum occurs where the acoustic round trip through the stack matches the contact time of the impact. For a contact set by the stack itself, one recovers the classical bar-impact condition of a stack-to-ball mass ratio of order one. Bar impact, however, predicts only a weak recovery, insensitive to the surrounding gas. For paper stacks, the recovery is large and no maximum is resolved when the stack is evacuated: the air trapped between the sheets, not the paper alone, makes the stack a trampoline.

Hengartner asked whether a pluriharmonic mapping between equidimensional complex Euclidean domains is locally one-to-one precisely at the points where its real Jacobian is nonzero. Naser had earlier claimed the conclusion in complex dimension two, and a recent theorem of Kalaj supplies a complete proof in that dimension. We prove the result in every complex dimension. The main ingredient is a singularity-theoretic characterization which is of independent interest: if $q\colon(\C^n,0)\to(\C,0)$ is a nonconstant holomorphic germ, then $dq(0)\ne0$ if and only if one (equivalently, every) phase hypersurface \( \{\Repart(e^{-i\theta}q)=0\} \) is a rational homology $(2n-1)$-manifold at the origin. The difficult direction combines local conical structure, Alexander duality, the Milnor fibration, and A'Campo's vanishing theorem for the Lefschetz number of local monodromy. The argument requires neither an isolated critical point nor a reducedness hypothesis. It follows that the real critical set of any pluriharmonic self-dimensional mapping agrees exactly with its local noninjectivity set. As a quantitative application, combining our theorem with Kalaj's boundary regularity theorem shows that, in every complex dimension, a quasiconformal pluriharmonic homeomorphism of the unit ball onto a bounded $C^1$-Dini domain is bi-Lipschitz.

Underwater robotic inspection depends on acquiring views that reveal task-relevant structure. For a structurally complex coral colony, recognising the target is only the starting point: the robot must select and execute a viewing motion suited to the inspection task. We present CoralPlan, a vision-language system that selects an observation skill from a current camera image and task text supplied by an episode manifest. A shared motion interface executes orbit, patch, or survey as target-relative trajectories; the remaining plan fields provide operator guidance. Observation completion requires target keeping and primitive-specific coverage, while joint success also requires selection to match the recorded reference. We evaluate this interface in 144 simulated episodes and 36 matched simulation-hardware pairs. In a clear-water pool with external target-reference poses, hardware observation completion reaches 77.8% and joint success reaches 63.9%. The experiments identify both reference-mismatched completions and incomplete observations after a matching skill selection. These results connect observation-skill choice to measurable underwater execution outcomes and identify where task-directed acquisition succeeds or fails.

Conventional silicon avalanche photodiodes (Si APDs) rely on highly doped regions to enable impact ionization and achieve internal gain. However, inefficient charge collection in these regions, combined with front surface reflection, degrades the external quantum efficiency (EQE) of these devices. In this work, we mitigate both loss mechanisms by integrating black silicon (b-Si) surface nanostructuring with Al2O3 induced carrier collection in an inverted-mesa Si APD architecture. This design confines the electric field and defines the multiplication region without requiring ion implantation. The resulting devices display near-ideal responsivity (at M = 1) across the UV-visible spectrum, with EQE exceeding 100% at 200-400 nm (peak ~130%) and exhibiting 92-100% at 400-700 nm. Avalanche gains of M ~ 4 at 80 V, M ~ 10 at 100 V, and M ~ 23 near breakdown at ~110 V are obtained. Dark current remains in the picoampere range up to ~80 V but rises to the nanoampere range near breakdown, narrowing the practical bias window. Capacitance-limited rise times of ~30-570 ns are measured for device diameters of 1-5 mm. The results demonstrate that the developed architecture enables APDs capable of detecting every single photon over wide range of wavelengths.

Motivated by models of mixed quantum-classical hybrid theories, we consider the representation of classical observables as Hilbert-space operators based on van Hove's unitary representation of contact transformations. This leads to a reformulation of classical mechanics where the phase space density satisfies the Liouville equation and, more importantly, classical observables are represented by van Hove operators that satisfy a commutation algebra isomorphic to the classical Poisson algebra of functions in phase space. Thus, this Hilbert space approach captures all the essential features of the phase space formulation of classical mechanics in that it reproduces its algebraic structure – something that is not achieved by other approaches such as the Koopman-von Neumann theory. As an important feature, the definition of appropriate physical states in this formalism imposes certain natural requirements on the phase of the classical wave function, which ensure correct expectation values for the observables. Furthermore, we discuss how to extend the approach to the hybrid van Hove theory of mixed classical-quantum systems and illustrate its application with the example of interacting classical and quantum oscillators. Finally, we focus on the question of how hybrid van Hove theory evades no-go theorems on mixed classical-quantum systems and point to the various assumptions of the no-go theorems that do not apply to our approach. This demonstrates that these no-go theorems are not universally applicable, contrary to their original claims.

We report an electron paramagnetic resonance (EPR) study of Dy${^3+}$ in CaWO$_4$ single crystals spanning dopant concentrations from 80~ppb to 247~ppm. The $g$ factors of the ground Kramers doublet at the $S_4$ point-group site, $g_\parallel = 7.22 \pm 0.01$ and $g_\perp = 5.45 \pm 0.01$, and the hyperfine constants of \ce{^{161}Dy} and $^{163}$Dy are determined with an order-of-magnitude improvement in precision over the only previous report. A crystal-field analysis constrained jointly by published optical levels and by the measured $g$ tensor identifies the ground doublet as the lower branch of an anticrossing between the near-degenerate $\|{\pm}11/2>$ and $\|{\mp}13/2>$ states, and explains why optical data alone left the $g$ values unconstrained. \add{The same parameter set, tested on the excited multiplets of the $^6H$ term, reproduces the ${}^6H_{13/2}$ crystal-field levels better than the optical parameters themselves.} The angular dependence of the linewidth in the $ab$ plane reveals broadening by random internal electric fields through the linear electric field effect. Calibrating these fields with the residual Er$^{3+}$ present in the same crystals yields the first electric-field coupling parameters of Dy$^3+$, $(B_{31}^2+B_{36}^2)^{1/2} = (34\pm5)\times10^{-6}$~cm/V, three times the Er$^{3+}$ value and the largest reported for a rare-earth ion in this host.The residual linewidth and the concentration-dependent lineshape asymmetry follow the quadratic field scaling expected for a second-order Stark shift, which escapes the inversion-site cancellation constraining the linear effect. Both couplings trace back to the $\sim 20$~cm$^{-1}$ gap to the first opposite-symmetry doublet, a consequence of the dense crystal-field structure of the $J=15/2$ manifold.

Achieving robust cross-embodiment generalization in imitation learning demands overcoming a critical representation flaw that inextricably entangles task semantics with hardware-specific visual geometry. We propose an interaction-centric framework that leverages the shared structure of two-finger grippers via a parameterized universal gripper abstraction, yielding a canonical gripper-frame representation. Given language and RGB-D observations, a VLM infers the subtask and grounds an interaction triplet (gripper, held, target), while SAM~2.1 tracks masks to reduce VLM queries. We design concise hybrid features that combine target/collision artificial potential fields for global guidance with segmented gripper-frame point clouds for local geometry, and use a Flow-Matching Transformer to predict smooth 7-DoF action chunks. Experiments in simulation and real-world tasks demonstrate that ours is the first imitation learning approach to simultaneously achieve competitive benchmark scores and extreme cross-embodiment/cross-viewpoint zero-shot sim-to-real transfer to completely distinct, heterogeneous robot platforms.

Auroral spectrographs such as the Auroral Spectrograph In Skibotn (ASIS) record hundreds of thousands of emission spectra, but only a few hundred can be labelled by an expert. To exploit the rest, we pretrain a 1D Vision Transformer with a masked autoencoder on 223,000 unlabelled spectra. Without labels, its representation recovers the emission-line intensity ratios that physicists use to diagnose the precipitating particles (R^2 0.91 vs. 0.77 for an untrained control) and, under one linear probe, classifies as well as 13 features designed by experts. Fine-tuned, the model outperforms the previous supervised auroral classifier on its own benchmark (macro-AP 88.5 vs. 77.8), reaches 0.870 mAP, and exceeds the same architecture trained from scratch by +0.159 with 10% of the labels; attribution shows that it uses both N2+ bands. Could an existing pretrained model replace it? Two astronomical spectral foundation models and a time-series model transfer according to their spectral window: SpectraFM, trained in the infrared, falls below the untrained control, whereas SpecFormer, trained in the optical, approaches in-domain pretraining without reaching it.

For distinct integers $a$ and $b$, an $\{a,b\}$-edge-weighting assigns $a$ or $b$ to each edge and labels each vertex by the sum of its incident weights. Such a weighting is proper if adjacent vertices receive distinct labels. We prove that, for every fixed pair of distinct integers, deciding whether a proper weighting exists is NP-complete even for simple cubic planar graphs. On planar multigraphs with $m$ edges, we give an exact $2^{O(\sqrt m)}$-time algorithm and, assuming the Exponential Time Hypothesis (ETH), exclude $2^{o(\sqrt m)}$-time algorithms even for simple cubic planar graphs. As a consequence, locally irregular $2$-edge-coloring is NP-complete on simple cubic planar graphs, admits a deterministic $2^{O(\sqrt n)}$-time algorithm on $n$-vertex graphs in this class, and admits no $2^{o(\sqrt n)}$-time algorithm under ETH. For maximizing the number of edges joining vertices with distinct labels, we give a deterministic efficient polynomial-time approximation scheme (EPTAS) on planar multigraphs, a polynomial-time $1/2$-approximation on multigraphs, and APX-completeness even on simple cubic graphs. Extending a partial $\{a,b\}$-edge-weighting to a proper one is NP-complete for every fixed pair even on simple cubic planar bipartite graphs, while it is polynomial-time solvable on trees. The hardness persists even when the prescribed edges form disjoint paths of length $6$ and all edges of each path have the same prescribed weight.

Can the spatial organization of cortical activity help transfer frozen image representations to fMRI prediction? We introduce FlatClip, which renders cortical activity as anatomy-linked flatmaps and extracts features with a frozen SigLIP2 encoder. Only a lightweight downstream probe is trained, without fMRI-specific encoder pretraining or fine-tuning. Across four resting-state benchmarks, FlatClip outperforms the evaluated ROI-level baselines on HCP and ADNI, but is weaker on PPMI and below the strongest voxel-level models overall. On NSD visual-fMRI decoding, restricting inputs to visual or NSD-provided task-active cortex improves recognition over whole-cortex inputs. Spatial perturbations reduce performance with both retrained and fixed readouts, and anatomy-linked arrangements outperform cortical-pixel permutations across three colormaps. Together, these results demonstrate a practical route for reusing image-pretrained features in fMRI and highlight cortical arrangement and input-region selection as important design choices. Code is available at https://github.com/OneMore1/FlatClip.

We present a theoretical study of pump-induced Faraday and Kerr rotation in a two-dimensional lattice of conducting disks. We propose a mechanism of Faraday and Kerr responses in which a circularly polarized pump directly induces a high-frequency Hall component $\sigma_{xy}=-\sigma_{yx}$ of the electron conductivity tensor at the probe frequency. Microscopically, $\sigma_{xy}$ originates from the third-order nonlinear response of the electron gas rather than from the real magnetic field created by solenoidal charge currents in the inverse Faraday effect. We show that the Faraday and Kerr rotation angles are significantly enhanced when the pump and probe frequencies are tuned close to the plasmon resonance of the disks, reaching $\sim 0.1^\circ$ per 1~kW/cm$^2$ of incident pump intensity in the terahertz range. This mechanism can explain recently observed giant pump-induced Faraday rotation in graphene disk lattices.

Composed image retrieval (CIR) aims to retrieve images that satisfy a user-specified modification while preserving relevant visual content from a reference image. Collecting target images for this purpose is costly, motivating zero-shot CIR methods that instead use target captions as supervision. However, target captions may omit source details that should be preserved. We therefore propose, Preserve-and-Compose Training, which complements target-caption supervision with visual evidence from the source image. PACT learns from image–text–text (ITT) triplets without target images or gallery updates, aligning composed queries with target captions while preserving source evidence through visual supervision. We further introduce Chord scoring, which combines target similarity with source-relative directional agreement in the frozen image space. Results across four ZS-CIR benchmarks show that combining target-caption supervision with source-image evidence leads to strong retrieval performance across datasets, backbone scales, and external galleries. The code is available on https://github.com/sehyunkwon/PACT.

Scaling studies typically represent training data by a single count of samples. For hierarchically and spatially structured data, however, the same number of samples can be drawn from few or many sources and distributed differently across the underlying domain. We therefore study data scaling as an allocation problem, separating unique sample count, source diversity, and spatial coverage. We study this decomposition in microscopic whole-brain histology, where a source is an individual brain, and a sample is an image patch at a specific spatial location. Across 93 controlled pretraining runs of a contrastive model that uses spatial proximity for supervision, we vary data allocation, compute, and model capacity over 11.6 million spatially anchored image patches from 21 human brains. Performance improves with more unique samples, broader spatial coverage, additional compute, and larger model capacity. At fixed sample count, distributing samples across one to 18 subjects produces no detectable improvement, even though representations generalize substantially better to subjects encountered during pretraining. Inter-subject variation therefore strongly affects generalization, but additional subjects provide no benefit when a fixed sample budget is distributed across more sources. These results establish sample count, source diversity, and spatial coverage as distinct axes of data scaling in spatially structured representation learning.

We prove that on the Sierpinski gasket, a prototype of compact fractal space, Dirac distributions arise as the weak limit of probability measures associated with sequences of high energy eigenfunctions of the Laplacian in a way that cannot happen in compact Riemannian manifolds.

Large-scale Digital Surface Models (DSMs) can be produced cost-effectively from satellite images via stereo-photogrammetry. However, the resulting 3D maps are often contaminated by noise, outliers, and voids. On the other hand, aerial LiDAR provides high-accuracy elevation measurements at a substantially higher cost. In this work, we study diffusion models conditioned both on photogrammetric DSMs and Pléiades imagery to refine vertically co-registered DSMs. We introduce a modified Stable Diffusion 3 architecture with a pruned text stream and a patch-wise normalization strategy, enabling stable training on LiDAR data and transfer from natural images to elevation maps. Experiments in French cities demonstrate that multimodal conditioning improves elevation accuracy, reducing Dense Urban RMSE from 6.00 to 3.45 m in the in-context cities and from 4.16 to 2.77 m in the held-out city of Bordeaux.

We present Light Field Primitives (LFP), a formulation for novel view synthesis that replaces the dense ray database with a compact set of differentiable primitives in the classical two-plane parameterization. Each primitive condenses a group of rays into one learned record, and its response to a query is governed by how closely that query belongs to the group. Rendering a camera ray then reduces to compositing all responses it elicits, and a scene can be optimized directly from posed images and rendered in real time with rays. Beyond its competitive performance on standard benchmarks, the main advantage of LFP is structural: its primitives reside directly in the 4D ray space, so optical and appearance effects that are already operations on the light field become behaviors of a single shared renderer. With minimal changes to that renderer, LFP supports multi-scale anti-aliasing, defocus deblurring with refocusing, rendering for fisheye cameras, and even transparent object reconstruction with ray refraction, matching specialized frameworks that devote substantial machinery to these effects.

Machine learning for scientific discovery is almost systematically data bound. Producing relevant high quality data, under budget constraints, is amongst the most promising ways to advance the field. Active learning (AL) offers promise wherever labelling requires expensive experiment, measurement, or simulation. Most existing tools cover only part of the data acquisition loop, and typically focus on either offline benchmarking or online deployment, but not both. We present ALF, a modular AL Framework that runs the full data acquisition loop via five modular components. One clear API for both settings: offline, against an existing dataset for controlled and reproducible experimentation; and online, against an oracle for acquiring new candidates in real-world deployments. ALF is open-source and available at https://github.com/instadeepai/alf.

Recirculating flows are central to transport phenomena, but their generation in potential flows requires additional physical mechanisms. We show experimentally that recirculation at vanishing Reynolds number can be produced at prescribed locations by coupling oscillations of a Hele-Shaw cell ceiling with an off-center passive obstacle. The net drift is well described by a potential flow model with no adjustable parameters. The resulting flows are localized around the obstacles and can be combined to generate programmable flow patterns at the microscale.

Automated EEG artifact removal may improve downstream analysis but can also alter predictive information. We benchmarked nine automated artifact removal methods against a common no-artifact-removal baseline for cross-dataset EEG age prediction. We introduce Signal Quality Index (SQI)-guided GEDAI, which leverages local signal-quality assessment to restrict correction to the channel–epoch pairs requiring intervention. Three deep neural architectures were trained on TUEG and evaluated without target-domain fitting on ds005385, LEMON, and TDBRAIN. Across this setting, GEDAI and SQI-guided GEDAI were the only methods with consistent gains over baseline in age prediction performance across all datasets and architectures ($\Delta$MAE $=-0.77/-0.64$ years, $\Delta R^2=+0.083/+0.072$, respectively). The remaining methods were neutral or detrimental on average ($\Delta$MAE $=+0.22\pm0.16$ years, $\Delta R^2=-0.023\pm0.016$ across methods). The two GEDAI-based methods achieved closely matched performance, while SQI guidance reduced the median modification ratio from $74.78\%$ to $42.80\%$. These findings show that curation benefits are method-dependent and establish SQI guidance as a more selective operating point, leaving more of the original EEG unchanged and limiting the potential loss of neural activity while retaining most of GEDAI's predictive benefit.

The zero-phonon line (ZPL) between the two excited-state manifolds 3F4 and 3H4 of the trivalent thulium ion in yttrium aluminum perovskite (Tm3+:YAlO3) lies within the telecom E-band at around 1451 nm and exhibits an optical coherence time of a few {\mu}s at cryogenic temperatures. To improve this time in view of potential applications, e.g. single-photon sources and optical quantum memory, we characterized the magnetic properties of the lowest Stark levels of these two manifolds. Varying the external magnetic field amplitude and orientation, we measured the enhanced nuclear Zeeman splitting by probing the field-dependent frequency detuning of side holes created by means of spectral hole burning, and we characterized the inhomogeneous line shift caused by the quadratic Zeeman interaction. Together, this allowed us to establish the spin Hamiltonians for the lowest Stark level of the 3F4 and 3H4 manifolds, and to predict optical clock transitions for particular orientations and magnitudes of the magnetic field

Current Audio-Visual LLMs (AVLLMs) struggle with reasoning over videos featuring multi-speaker dialogues. In such videos, resolving "who says what" is crucial, which necessitates trimodal (text-audio-visual) binding. Motivated by these challenges, we systematically investigate how this trimodal binding is achieved in AVLLMs. Specifically, we identify emergent symbolic trimodal binding mechanisms in AVLLMs that utilize modality-specific symbolic variables. By encoding auditory and visual components into symbolic variables-capturing temporal utterance sequences and spatial entity coordinates, respectively-the model establishes cross-modal linking within this abstract space. Crucially, we reveal that when trimodal binding fails, the breakdown predominantly stems from misaligned audio-visual connections. To overcome this bottleneck, we introduce an audio-visual prompting method utilizing an off-the-shelf Active Speaker Detection (ASD) model. By simply overlaying visual bounding boxes on active speakers, this training-free approach yields immediate performance gains across four conversation-centric benchmarks. Moreover, lightweight fine-tuning of fewer than 300 steps on these ASD-prompted-videos extends these gains to three general AV benchmarks, suggesting the generalizability of our method.

How can a rule inferred from examples reach someone who never saw them, without a shared code? Patients with severe aphasia do it by gesture or sketch. One network sees worked examples and emits eight invented symbols; a second, blind to them, applies them to a new input. Rewarded for the second's success, the first learns a code carrying rules to three-step transformations training never presents, which new learners acquire. Like invented human languages, the code has two regimes: under reward alone the speaker drifts to one message, as human languages lose words under plain transmission; expressive pressure keeps messages differentiated. Success on new rules tracks how much the message says about the rule, not how varied messages are. The learning signal shapes the code: reward sorts many rules under few fixed labels; the listener's error gradient gives each rule a region of similar messages, telling rules apart far better.

Data quality research has usually treated data as an input that is stored, processed, and validated. In AI-driven software-intensive systems, data also shapes model behavior, evaluation, and lawful use. Empirical evidence remains limited on how practitioners define, assess, and manage quality under these conditions. We interviewed 16 practitioners from nine organizations and analyzed the transcripts using reflexive thematic analysis and developed six themes from participants' accounts. In AI systems, traceability shifted from modular debugging to attributing model behavior, while using models as quality assessors introduced circularity. Agent context and memory became data objects, and synthetic and pseudo-labeled data made authenticity a quality concern. In foundation-model development, lawfulness became a gate for training data, while representativeness was judged through coverage of situations in which the system must behave safely. Prior ML research examines many of these problems separately. Our study provides a practitioner-grounded account of how they are encountered together as an engineering and organizational concern. We also interpret five recurring conditions as helping explain how the themes relate to reduced trust in data and AI outcomes. We synthesize these findings through lifecycle assurance: a conceptual framing focused on producing evidence that data can support a specific AI claim when its influence may be embedded in model behavior, model-based judgments, or agent actions.