Papers

Studying real-time degradation in perovskite semiconductors remains an important challenge. Conventional methods often cannot separate interfacial and bulk processes or rely on destructive analysis. Here, we demonstrate multiwavelength excitation photoluminescence (PL) as a nondestructive approach to distinguish interface- and bulk-dominated recombination losses. Using phenethylammonium tetrafluoroborate (PEABF4) treatment as a model, we use PL to show that PEABF4 passivates the electron transport layer (ETL)/perovskite interface, reducing interfacial recombination losses from fullerene deposition. However, upon thermal aging, PEABF4-treated stacks degrade more rapidly, with recombination losses broadening from the ETL/perovskite interface into the absorber. Diffusion-recombination simulations support interpreting excitation-wavelength-dependent PL decay signatures consistent with changes in surface recombination velocity and bulk lifetime, providing a framework for depth-sensitive comparisons during aging. Collectively, these results further establish multiwavelength excitation PL as a platform for studying recombination losses during aging of perovskite films.

Conventional transmission electron microscopy (TEM) records two-dimensional (2D) projections of three-dimensional (3D) objects, complicating the inference of nanoparticle morphology. This limitation is investigated for quasi-spherical iron oxide nanoparticles (IONPs) imaged after deposition on a TEM grid and after cellular internalization and immobilization in resin. A statistical inverse-reconstruction method is developed in which synthetic 3D particle populations are randomly oriented, projected onto a TEM-like plane, and compared with the experimental equivalent-diameter and axial-ratio distributions. Models of increasing geometrical complexity are considered, from monodisperse prolate spheroids to size- and shape-polydisperse prolate populations and triaxial ellipsoids. Size polydispersity is required to reproduce the projected equivalent-diameter distribution, whereas relaxing axial symmetry through a triaxial description enables simultaneous reproduction of the size and axial-ratio distributions. The inferred population represents a statistically compatible 3D morphology rather than a unique particle-by-particle reconstruction. Its physical relevance is assessed through a macrospin model incorporating the corresponding shape-anisotropy contributions. Application to magnetic fluid hyperthermia reveals a pronounced sensitivity of the predicted heating response to particle geometry and polydispersity. The approach provides a general framework for connecting conventional TEM measurements with morphology-dependent properties of nanoparticle ensembles.

Iterative Linear Quadratic Regulator (iLQR) and Differential Dynamic Programming (DDP) are attractive trajectory-optimization methods because their backward recursions provide both nominal trajectories and local time-varying feedback laws. Hard state and input constraints, however, can drive solutions toward constraint boundaries where little corrective authority remains during closed-loop execution. This paper presents Box-iLQR, a log-barrier iLQR method for box-constrained states and controls, with emphasis on the resulting feedback policy. Barrier gradients bias nominal trajectories away from constraint boundaries, while barrier curvature enters the Riccati recursion that determines the feedback gains. We show that control-barrier curvature smoothly attenuates feedback as an actuator approaches saturation, while state-barrier curvature propagates backward through the value function to shape feedback before future state constraints are encountered. This motivates retaining a finite barrier rather than always driving the barrier parameter toward zero, thereby preserving state and actuator margin. Comparisons with CL-DDP and ALTRO are presented on pendulum, cart-pole, Acrobot, and car-parking problems. Monte Carlo experiments demonstrate improved closed-loop behavior with finite barriers, while a six-input MuJoCo fish example demonstrates data-driven Box-iLQR using dynamics Jacobians estimated from simulator rollouts. Box-iLQR achieves convergence comparable to, and in some cases better than, the baseline methods. Moreover, the feedback policy obtained with a finite barrier ensures constraint satisfaction even under execution uncertainty.

Health technology assessment often aims to balance evidence across multiple treatments and outcomes and assess treatment benefit-harm profiles. Multivariate network meta-analysis (mvNMA) is well suited to this task, as it synthesizes evidence across treatments and outcomes. However, the potential of mvNMA has been overlooked, leading decision-makers to rely on outcome-specific NMAs that ignore outcome correlations. Based on independent NMA models, previous work proposed methods such as multivariate P-scores or spie charts to assess treatment benefit-harm profiles by yielding across-outcomes treatment hierarchies. In this article, we propose an mvNMA-based framework for assessing treatment benefit-harm profiles. Our framework uses a hybrid-correlation mvNMA model and incorporates DuMouchel priors for treatment effects, enabling estimation of all treatment effects even when outcomes are missing for some treatments. We then extend common NMA ranking metrics, including SUCRA, the probability of being best, and median ranks, to the mvNMA setting. Based on these metrics, we adapt the VIKOR algorithm for mvNMA. This deterministic multicriteria decision method quantifies overall and worst-case treatment performance and, by balancing these aspects, produces an across-outcomes ranking while identifying treatments offering the optimal compromise between these two aspects. To facilitate use of our framework, we developed the R package mvnma. We applied our framework to two clinical datasets: one comparing eleven physical therapy treatments across two outcomes and another comparing nine antidepressants across five outcomes. Our framework provides an alternative strategy for assessing treatment benefit-harm profiles across multiple outcomes and encourages multivariate approaches in applied clinical and policy settings.

Methods for assessing language-model agents include rule checkers over logs, analyses of skill coverage and composition, support checkers, prefix monitors, execution gates, and rare-event estimators. Each observes a different part of a run and makes a claim of a different strength, and no common account says how these claims combine or what they leave unchecked. We give one, built from a logic, information fragments, and an assurance ledger. Requirements are formulas of a two-tier logic: an outer finite-trace temporal logic over recorded events, and an inner logic of standing over the argument structure that the agent's recorded context supports at a decision. A rule violation and an action taken on withdrawn support are thus formulas of one language. The information available to an assessor, to the agent, and to an execution gate defines fragments of that language; each checking method decides one fragment and returns a typed claim: exact, on a named test suite, a risk bound, or descriptive. The ledger merges the evidence and classifies every obligation as established, addressed but not established, or unaddressed. On 456 released $\tau^2$-bench telecom trajectories, the benchmark oracle flags 231 runs; adding a rule checker, a support stand-in, and a prefix-monitor baseline raises the union to 391, 401, and 407, each contributing flags the others miss, and a gate makes one prohibition exact on a gated deployment. The 49 unflagged runs and the unmet or unaddressed obligations form the residual. Discovery makes unknown requirements explicit, and new or stronger checkers then reduce it.

We study formal matrix rings with the involution property, that is, rings in which every element can be represented as the sum of an invertible element and an involution. After recalling the construction of formal matrix rings via Morita contexts, we describe involutions in triangular formal matrix rings and establish a criterion for upper triangular rings of order two. We then prove that a formal matrix ring of arbitrary finite order has the involution property whenever all of its diagonal rings have this property. Several examples and counterexamples are included to illustrate the obtained results.

Operator learning methods such as DeepONets and FNOs often struggle with PDE families featuring sharp interfaces, heterogeneous coefficients, and localized multiscale structures. We introduce a partition-of-unity (POU) mixture-of-experts framework for localized operator learning, in which geometry-aware gating networks produce smooth spatial partitions which blend the contributions of local expert networks. Our main contribution is HiRefPOU, a residual-style hierarchical POU architecture for DeepONets that organizes localized representations through nested parent-child partitions while preserving global continuity. We also show that the same POU principle can be incorporated into Fourier Neural Operators to introduce spatial adaptivity without modifying the underlying spectral layers. On heterogeneous Darcy and reaction-diffusion benchmarks, HiRefPOU achieves substantially lower error than global DeepONet and static POU-MoE baselines, while the broader operator-learning experiments show that the benefits of localization depend on the PDE structure and the chosen neural-operator backbone. The learned partitions are interpretable and align with interfaces and regions of rapid solution variation. These results show that explicit geometric localization can improve both accuracy and interpretability in neural operator learning.

Laboratory determinations of the Newtonian gravitational constant $G$ disagree by more than their quoted uncertainties allow. We examine the sixteen determinations comprising the CODATA-2018 archive using a proportional covariance-scale model that retains the full reported covariance matrix. Because the 2026 NIST torsion-balance redetermination shares apparatus lineage with the BIPM Mark II balance, our primary historical fit excludes the BIPM-14 determination, giving $\hat{\lambda} = 2.967$ for the remaining fifteen measurements. Evaluated against this historical model under the declared uniform covariance-scaling convention, the correlated four-configuration NIST result is highly unlikely under the archive's reported, unscaled covariance but is statistically unremarkable once the historically inferred extra dispersion is applied ($Q_\lambda = 4.840$; finite-sample $F$ tail $p = 0.382$); a mean-only scaling sensitivity leaves the vector discrepant. Retaining the complete archive also gives a comparable result. We further examine four matched free-minus-servo torque differences reported by NIST. They are consistent with a single common absolute torque offset across the four tested configurations, $\widehat{\Delta N} = 1.918 \pm 0.215$ pN m, and poorly described by an equally simple fractional-scale alternative under the stated independent-Type-A treatment, with a chi-square contrast of $\Delta\chi^{2} = 11.041$ between the two torque models. The paper provides a covariance audit of the historical archive, a lineage-qualified consistency check, and an experimentally testable diagnostic for dual-mode torsion balances, without proposing a revised value of $G$.

People write and sketch while speaking to explain, organize, and develop content together. Inspired by these practices, we investigate how voice agents can use a canvas alongside speech in multi-turn conversations with users. We conducted a two-part formative study: an observational study of how pairs coordinated speech and boardwork, followed by a design workshop that informed a design space for speech-canvas interaction with voice agents. Building on these insights, we developed VoCa, a voice agent that coordinates speech with visual object creation, annotation, and attention guidance. A five-day deployment with 18 participants examined usability, experiences of speech-canvas interaction, patterns of use, and desired improvements. Participants' experiences highlighted opportunities for speech-canvas interaction in learning, work, and daily life, alongside challenges in coordinating what agents say and show in ways users can follow and influence. These findings inform how voice agents can use a canvas alongside speech in conversation.

Automated neurodevelopmental assessment increasingly combines computer vision, speech, eye tracking, physiology, and machine learning, yet multimodality and discrimination do not establish construct validity, clinical usefulness, or appropriate interpretation of behavioral variation. We propose a testable architecture for neurodiversity-aware multimodal affective computing in which population-relative deviation is not treated as sufficient evidence of adverse functioning. This conceptual article reports neither a new participant-level study nor a trained predictive system, and no predictive or clinical superiority is claimed. The framework keeps observable signals, modality quality and availability, context, population-relative information, pooled within-person references and, when sufficiently supported, context-conditioned within person references, and predictive uncertainty distinguishable throughout inference. It defines structured decision-support outputs and auditable requirements for target validity, temporal integrity, contextual leakage, missing-modality robustness, calibration, subgroup evaluation, interpretability, privacy, and human decision boundaries. The contribution is architectural rather than algorithmic: it specifies empirically testable constraints that can be instantiated with different multimodal learning methods. Autism provides the principal motivating evidence base, without assuming unchanged transfer to other neurodevelopmental conditions.

We study the invertible one-form global symmetry of a (2+1)-dimensional topological quantum field theory (TQFT). A universal description of this symmetry “peels off” all the Abelian anyons from the rest of the TQFT. This allows us to study this symmetry and some of its consequences, independently of the details of the rest of the system. We then describe these Abelian anyons using $\mathrm{U}(1)$ gauge fields with Chern-Simons terms and couple them to a scalar field, thus constructing a particular non-topological theory. This new theory, our transition field theory, exhibits a phase transition from the original TQFT to another phase. The latter is either gapless or another TQFT. This discussion generalizes and extends our earlier work (arXiv:2603.00245). In particular, we give a more detailed description of the global symmetries and their fates in the various phases.

Recently proposed Drifting Models shift iterative distribution refinement from inference to training, enabling effective one-step generation. However, their performance on complex image datasets depends strongly on the representation used to construct the drifting field: pixel-space drifting performs poorly, whereas pretrained feature spaces substantially improve sample quality for reasons that remain unclear. We trace this gap to the discriminative geometry of the representation, which determines sample weighting in kernel density estimation (KDE) and, consequently drift. We introduce persistent representation learning, which continuously learns a more discriminative representation geometry as the generator evolves across batches. We further establish a current-step gradient equivalence between the KDE ratio loss and drift regression loss under matched conditions, connecting density-ratio-based generator optimization to empirical drifting and motivating direct control of the drifting velocity. Across multiple datasets, our method learns effective discriminative representations directly from pixels and reduces FID by approximately $82-95\%$ over the original pixel-space Drifting Models, without pretrained encoders. Adapting pretrained representations and applying velocity clipping provide further gains.

Let X be a smooth projective curve of genus g >= 2 over an algebraically closed field K, and suppose that Aut(X) contains a cyclic subgroup G of order N > 2g - 2. The cases N >= 2g + 1 are known. We treat the two boundary values N = 2g and N = 2g - 1. In the tame case we obtain an explicit finite list of ramification signatures. In odd characteristic we prove that the Sylow p-subgroup of G has order p and determine all wild boundary cases by a direct ramification analysis; the resulting curves are Artin-Schreier-Kummer fiber products. In characteristic two we prove that the 2-part has order at most four and classify all wild boundary cases, including Artin-Schreier-Witt normal forms for the C4 and C12 cases.

Memory is essential for long-horizon robotic manipulation, where successful actions may depend on past events that are no longer recoverable from the current observation. As episodes grow longer, however, retaining the full history becomes increasingly costly, creating a fundamental scalability challenge for memory-augmented policies. Existing approaches address this challenge by either storing selected past observations in a bounded memory bank or compressing interaction history into a fixed-size parametric state through Test-Time Training (TTT). Yet these formulations do not explicitly distinguish between historical information that can already be recovered from the policy's current context and information that must persist beyond it. We introduce TTT-RM, which repurposes TTT as Residual Memory, using fast weights not to generically compress history but to complement a bounded memory bank by preserving task-relevant historical information that cannot be recovered from the policy's current context. Concretely, TTT-RM learns a history decoder that reconstructs historical representations from the current observation and retrieved memory. The resulting reconstruction residual captures what this context fails to explain and serves as the learning target for TTT. The TTT slow weights are optimized against this residual target so that online fast-weight updates learn to encode complementary historical information over time. The fast-weight state is then queried to produce a residual memory representation that conditions action generation. Extensive experiments on memory-intensive simulation benchmarks and real-world tasks show that TTT-RM consistently improves across multiple memory designs, outperforms diverse baselines, and supports sustained execution on a three-minute, eight-stage stowing task.

Federated learning enables medical institutions to train a global model without sharing data, yet feature heterogeneity from diverse scanners or protocols remains challenging. Existing representation-based methods face two limitations: 1) Incomplete Contextual Representation Learning: single-layer or coupled representations overlook multi-level structural cues and entangle regional semantics with boundary details. 2) Layerwise Style and Aggregation Biases: domain-specific style discrepancies across intermediate layers degrade prototypes, while aggregation that overlooks client distribution shifts can further amplify bias. We propose FedBCS+, federated decoupled contextual alignment with style-purified aggregation. We employ Frequency-domain Style Recalibration (FSR) in prototype construction to decouple content-style representations and extract style-purified prototypes. Built upon these purified features, Decoupled Contextual Prototype Alignment (DCPA) explicitly decouples multi-level features into semantic and structural prototypes and aligns regional semantics and fine-grained anatomical structures separately. Style-purified Semantic Prototype Aggregation (S2PA) measures each client's purified prototype divergence from the global consensus and adaptively reweights aggregation toward under-represented clients to reduce consensus bias. On five heterogeneous medical segmentation benchmarks spanning histopathology, MRI, ultrasound, and colonoscopy, FedBCS+ achieves the highest mean Dice among the compared methods. A convergence analysis further characterizes how aggregation and alignment affect the optimization bound.

A verifier for an agent faces rules of two kinds: the ones a fixed check can settle and the ones that require a judge. A team that derives its own checks fixes that split up front. Where the requirements come from outside, as in finance, healthcare and law, the agent enforces rules it did not write, so the split falls to runtime, recurring for every predicate of every rule on every action at a rate no reviewer can audit. Every escalation scheme assumes a model can make that decision itself, that it is self-decidable. Across six corpora, including the EU AI Act, FINRA guidance and a deployed credit agent, we collect roughly 22,000 labels from four models built by three labs. They agree almost perfectly where the answer is obvious and collapse on regulatory text; their errors run in opposite directions, so no model can be trusted as the conservative choice; and on the deployed agent's own rule-set they err together, over-claiming that a fixed check will do, the direction that never gets escalated. We introduce CoVer (corroborate-then-verify), which treats unanimity as a nomination, admitting a predicate only when the check synthesized for it survives intervention, reading fields the agent cannot write and holding under deterministic rewording. That gate rejects most of what corroboration wrongly admits, at a cost in coverage we report rather than tune away. The obvious alternative, agreement with a reference judge, certifies nothing: it climbs from 30% to 77% across calibration bands while the genuinely decidable share does not move, because a judge drawn from the population under indictment ratifies the blind spot it shares. Self-decidability is not a capability to elicit from a model but a boundary the verifier must construct.

The decametre radio sky ($\sim 10$-$30$ MHz) remains largely unexplored due to strong ionospheric distortions and severe radio-frequency interference. Despite its scientific potential, including access to steep-spectrum synchrotron emission from aged electron populations and coherent radio emission from planetary and stellar magnetospheres, sub-arcminute imaging at these frequencies has only recently become feasible with instruments such as the Low Frequency Array (LOFAR). In this work, we present the first data release (DR1) of the LOFAR decametre Sky Survey (LoDeSS), which maps the northern sky above a declination of $20^\circ$ between 15 and 30 MHz. LoDeSS reaches an angular resolution of $\sim45^{\prime\prime}$ and typical noise levels of $\sim11.5$ mJy beam$^{-1}$ at a central frequency of 23 MHz. Observations were obtained with the LOFAR Low Band Antenna (LBA) system using a multi-beam strategy, with typical integration times of $\sim5$ h per pointing. They were carried out at night to reduce the impact of ionospherically reflected radio-frequency interference. Data quality varies across fields as a result of the ionospheric conditions and the fraction of data flagged due to radio-frequency interference. This first release covers $\sim15\%$ of the full survey area and consists of fields originally selected around exoplanet candidates for radio auroral emission. The products released to the community include primary-beam-corrected images in Stokes I and V, and a source catalogue for fields with reliable flux density scales. This release demonstrates that LOFAR enables sensitive, wide-field imaging in the decametre regime with unprecedented resolution, opening new opportunities for studies of pulsars, exoplanetary systems, galaxy clusters, and other low-frequency phenomena.

In computational neuroscience, task-trained recurrent neural networks (RNNs) are commonly used as a testbed for exploring the space of recurrent circuit solutions compatible with a particular function. However, these networks are subject to inductive biases that may be misaligned with those of biological circuits, which operate under various efficiency and robustness constraints. Here, using a minimal stimulus recall task, we systematically characterize how the solution space of recurrent neural networks is shaped by the desiderata of weight efficiency, activity efficiency, and robustness to noise. We show that weight efficient solutions exhibit a low-rank bias that aligns with the inductive biases of task-trained networks and favors dynamical motifs involving slow, persistent dynamics, whereas activity efficiency encourages high-rank solutions that rely on transient amplification. Moreover, we find that activity efficient solutions exhibit a strong dissociation between high-variance modes of activity and those that causally influence network output. Unlike in static feedforward settings, we demonstrate that noise robustness and activity efficiency are generally in opposition, rather than synergistic. Using these findings, we propose a normative account of misaligned coding phenomena recently identified in premotor cortex.

Diffusion and flow models provide expressive policy classes for online reinforcement learning (RL), enabling multimodal behaviors and improved performance. However, training these policies remains challenging: the critic specifies the desired policy as an unnormalized Boltzmann density but does not provide direct samples from it. Many existing methods rely on importance sampling to construct training signals, which can suffer from high variance, increasing computational cost and destabilizing training. We propose Score-Calibrated Flow (SCF), a simple and efficient algorithm for training generative models to sample from unnormalized densities without importance sampling or backpropagation through the sampling trajectory. We learn the desired flow by enforcing self-consistency, bypassing target posterior mean estimation. By jointly exploiting the prescribed target score and the structure of flow matching, we establish these self-consistency requirements as score-calibrated optimality conditions, first for the terminal density and then for the trainable velocity field. We prove that their unique solutions are, respectively, the target density and the ideal flow model that conditional flow matching (CFM) would recover if target samples were available. We formulate the velocity condition as a fixed-point equation and exploit its conditional-expectation structure to construct a stop-gradient objective for enforcing it. The resulting training procedure retains the scalable sample-interpolate-regress structure of CFM despite the absence of target samples, using endpoints generated by the current flow. For online RL, the critic gradient supplies the target score at the generated actions, yielding a direct approach to actor training. Experiments on RL benchmarks demonstrate that SCF matches or improves upon state-of-the-art generative-policy baselines, while substantially reducing training time.

Operator eigenvalue problems appear throughout science. Classical methods usually discretize the operator into a matrix and then solve the resulting matrix eigenvalue problem. This works well in low dimensions, but fine grids quickly become expensive in both memory and computation as the dimension grows. Neural network based solvers avoid storing these large grids, but recent state of the art neural methods can require hundreds of thousands of training steps and may struggle to find the desired eigenvalues. We show that the two approaches can help each other. A coarse finite difference method (FDM) calculation acts as a cheap numerical model of the operator spectrum. We use the approximate eigenvalues as fixed shifts during the training of the neural solver, as they only need to locate the relevant part of the spectrum. We also introduce Stabilized Inverse Power Method Neural Network (SIPMNN), a more stable training procedure for higher-dimensional problems. Across five test problems at $d=10$, the combined approach is more accurate overall than the tested fully neural alternatives while using eight to ten times fewer iterations.

We develop a theory of stabilizer codes whose stabilizer groups consist of commuting qudit Paulis, oscillator displacements, and/or planar-rotor displacements. We consider codes whose stabilizer groups form generalized lattices in quantum phase space, a condition that guarantees a finite logical dimension. We construct oscillator-rotor, rotor-qudit, and oscillator-rotor-qubit codes that cannot be decomposed into separate subsystems by generalized Clifford transformations. We also introduce an oscillator-qubit code arising from a partially lifted Golay code, where Construction A is applied to half of the Golay code coordinates. We derive analytical forms of logical operators from the symplectic dual lattice, determine Clifford-Gaussian gates from lattice symmetries, show that the code dimension equals the covolume of the stabilizer lattice, and organize the syndrome subspaces into a vector bundle. Our technique builds on earlier non-commutative geometric results by Rieffel, but it can often be interpreted simply as concatenating a given stabilizer code with a Gottesman-Kitaev-Preskill (GKP) code, applying conventional lattice-theoretic results, and unconcatenating. We formulate distances, optimal decoding, and quantum weight enumerators, recovering the Pauli, qudit, and GKP versions together with known MacWilliams identities. We construct finite-energy codewords whose encoding is an isometry up to an error exponentially small in the inverse damping parameter.

Agents are entering finance, healthcare and law, sectors where a violation leaves no lexical signature and carries real penalties. Whether an omission is material, or a disclosure sufficient, depends on what the response left out. Probing such an obligation means finding responses one minimal edit from flipping compliance, and every probe costs a generation and two adjudications, so the binding constraint on regulatory red-teaming is sample efficiency, not volume. We introduce Penumbra, an adversarial search that walks from a verified anchor under an expanding edit budget until a two-evaluator committee changes its verdict, and emits the two adjacent responses that straddle the change. Allocation is adaptive, and the objective is coverage of the defeat surface: distinct (obligation x defeat mode) cells resolved per candidate. At matched budget, adaptive allocation reaches uniform allocation's full-budget coverage on 59% of the candidates; at equal records it covers 1.43x the defeat modes of naive enumeration, and the gain is confined to the axis it targets. On 60 screened obligations of a financial advisory constitution, Penumbra returns 144 pairs, each a compliant and a violating response that the committee places on opposite sides of the boundary, differing by a handful of words where a model asked for both directly produces texts sharing almost nothing. A second constitution, for clinical triage, reproduces this on 18 obligations: 49 pairs at the same tightness, with the same modes hardest. Pairs like these show where an obligation's own terms stop deciding, which is what an agent deployed under it must be tested against, and the search finds them at a cost that scales with the boundary, not the text.

While standard Hamiltonian and Lagrangian methods are powerful for unconstrained classical systems, they are not directly suited to constrained ones, despite the existence of well-established treatments in the literature. In this paper, we present an alternative approach to deriving a generalized Hamiltonian and Lagrangian framework, based on the revised Barcelos-Wotzasek (BW) symplectic algorithm. This method enables a consistent and successful construction of the standard Lagrangian and Hamiltonian descriptions for constrained systems.

Retrieval-Augmented Generation (RAG) is typically evaluated under the implicit assumption that the underlying knowledge base (KB) is clean, leaving the behaviour of RAG systems under realistic KB degradation poorly characterised. We introduce RAGStress, a controlled evaluation benchmark for stress-testing RAG systems under systematic KB corruption. The benchmark pairs four naturalistic corruption types (factual corruption, numeric typo, relevance poisoning, and contradiction injection) with three severity levels (subtle, moderate, and obvious) over a single-KB, metadata-filtered experimental design built from 57 MMLU subjects and 182,546 documents. Across 52,500 model-question-condition evaluations, RAGStress reveals that clean retrieval can mask robustness differences, semantic-fidelity corruptions are substantially more harmful than signal-utility perturbations, no-retrieval accuracy does not predict corrupted-retrieval robustness, and mixed-KB accuracy should not be treated as worst-case robustness. We document the benchmark's intended use, supported claims, and limitations, and provide an artifact bundle including generation scripts, corruption prompts, metadata schema, and evaluation code. RAGStress is intended as a controlled stress test for RAG robustness under KB corruption, not as a general model leaderboard.

Self-supervised learning (SSL) is standard for speech representation learning, but mainstream models are designed around single-speaker audio, limiting their usefulness in multi-speakers scenarios. We present SepRQ, an open-source SSL framework that replaces masked prediction with a pseudo-source-separation objective over frozen random-projection codebooks. By adopting a novel mask-free, multiresolution approach, SepRQ achieves state-of-the-art performance in Speaker Diarization and Speech Separation on the SUPERB benchmark, surpassing WavLM and other cocktail-party derived SSLs at both Base and Large scales, while requiring only 85.68M inference parameters. SepRQ also demonstrates strong performance across target-speaker tasks requiring enrollment (such as Target-Speaker Automatic Speech Recognition), and on the challenging multi-domain DIHARD 3 diarization dataset. Notably, we report strong separation capabilities on three-speaker mixtures (WSJ0-3Mix), where current SSL literature struggles. While cocktail-party SSLs remain scarce and closed-source, limited to C-HuBERT and the enrollment-based SA-WavLM, we open-source SepRQ to the community.

Time integration has been the cornerstone of physics-based animation that enables the simulation of complex interactions between rigid and deformable objects, including the motion of hair. Despite recent advances with optimized time integration that enabled thousands of hair strands to be simulated in real time, achieving the same performance on commodity hardware remains infeasible due to the computational demands of resolving complex dynamics and interactions between thousands of individual strands. With the rise of learning-based techniques, the offload of time integration to neural networks helps to achieve significant performance gains, making these approaches suitable for real-time applications such as gaming and virtual avatars. However, state-of-the-art neural techniques tend to produce less physically plausible motion and oftentimes fail to generalize to out-of-distribution scenarios. Inspired by classical time integrators, we design a neural counterpart that mirrors their input-output formulation – taking previous hair states, material stiffness, and collision geometry as the inputs for the neural time integrator, which is then trained via a self-supervised, simulator-in-the-loop method with randomized unrolling horizons. By formulating training in each strand's local coordinate frame, we obtain a network that generalizes across multiple dimensions, including hairstyle, material property, body motion, and body type. Our method inherits the benefits of a strand-based neural simulator, and hence is density-independent, lightweight, memory-efficient, and performant. Our neural hair integrator produces stable long-horizon rollouts and can be naturally extended to support quasi-static simulation simply by resetting hair states.

Generative AI enables authors to produce more manuscripts at lower cost. But evaluation by expert editors and referees remains scarce. In this context, the value of an author's submission rights can increase. Major research repositories and journals are reacting by imposing submission quotas and fees. Non-transferable author quotas transform submission of a manuscript into a revealed-priority decision. Authors construct a portfolio of their highest-ranked manuscripts. A common fee charges an author for every manuscript they submit. Both methods have drawbacks. Quotas do not flexibly allocate scarce review capacity to authors who have unusually strong realised candidate sets. Fees can generate heterogeneous submission thresholds that may be systematically related to an author's baseline portfolio quality. In this paper, I derive the condition under which submission quotas yield weakly higher mean reviewed value than fees, given an equal expected review load. When project values are bounded and candidate manuscripts are abundant, an equal-load fee can price out some authors entirely.

Answering questions over imperfect tables requires handling errors that can affect the answer. We investigate two challenges for large language models (LLMs): whether error discovery depends on where errors appear in a table, and whether providing their locations is sufficient for accurate question answering (QA). Using human-reviewed instances from RADAR-T, we conduct controlled studies across three LLMs by varying row order and comparing original, error-marked, and repaired tables. First, reordering rows changes error discovery even when the table contents and gold answer remain unchanged. During direct inspection, LLMs are more likely to discover all rows containing relevant errors when these rows appear later in the table or are grouped more closely together. Second, providing verified error locations alone is insufficient for accurate QA: with code execution, accuracy on repaired tables exceeds that on error-marked tables by 39.0-59.1 percentage points across the three LLMs. As a practical application of these findings, we combine error discovery across shuffled table views with explicit guidance for verifying and handling the reported errors in a simple workflow, Geometry-Balanced Discovery and Intervention (GBDI). On RADAR-T, GBDI improves QA accuracy by 3.8-18.5 percentage points over a code-agent baseline across five LLMs (paired 95% confidence intervals exclude zero for four), at the cost of additional inference. These results highlight the importance of both reliable error discovery and effective error handling in QA over imperfect tables. Code is available at https://github.com/645-t/GBDI-ICLR-2027.

Multi-agent debate can improve reasoning, yet often fails to beat simple majority voting. We argue that successful debate requires two distinct mechanisms: proposal supply must surface a correct answer, and readout must identify that answer when voting misses it. We formalize the first requirement through recoverable headroom, which measures cases where a correct proposal is available but the majority answer is wrong. For the second, we develop Latent Verification Debate (LVD), an accounting model in which candidate proposals receive answer-specific verification evidence before final generation. Controlled fixed-proposal interventions estimate this latent effect in equivalent peer-support units and show that correct evidence changes answer probabilities and generated decisions while proposal supply remains fixed. To improve proposal supply, we construct societies from neural-thicket agents using labeled and label-free coverage objectives. Across two backbones and matched-budget reasoning benchmarks, coverage-selected societies increase complementary proposal supply and improve aggregate accuracy in repeated stochastic evaluations. Round-level controls further show that interaction provides gains beyond applying the same finalizer directly to the initial proposals. These results identify proposal coverage and truth-sensitive evidence use as complementary conditions for debate to outperform voting. Code is available at https://github.com/Wang-ML-Lab/when-debate-helps.

This paper refines randomized block coordinate descent by introducing a coordinate-level batch-sampling distribution characterization, yielding a fine-grained randomized batch-sampling coordinate descent framework, which applies to any block coordinate descent method under an arbitrary static stochastic sampling rule for solving large-scale linear least-squares problems. Based on this framework, we derive new linear convergence rate bounds that are invariant to the column scaling of data matrices. Theoretical analysis and numerical experiments validate that the new bounds are significantly sharper than existing results and better aligned with the practical convergence rate. Furthermore, we proposed a learning-based sampling strategy by parameterizing the batch-sampling distribution and optimizing its parameters through gradient-based training. Preliminary numerical results on computed tomography (CT) reconstruction problems demonstrate that the resulting method converges significantly faster than variants using other fixed sampling rules, and outperforms tested adaptive coordinate descent methods and full gradient descent in computational efficiency, confirming that learned sampling distributions can effectively capture problem-specific structure to accelerate convergence within the framework.