Papers

Qualitative comparison figures are central evidence in computer vision papers, and vision-language models (VLMs) are increasingly used to judge them. Yet existing benchmarks score only scalar quality or overall preference, so a judge can be rewarded for picking the preferred image for the wrong visual reason. We introduce VisionQ, the first benchmark built from peer-reviewed CV comparison figures that grounds every judgment in a named visual criterion: each question states the criterion, and a judge is credited only when it selects the output the authors identify as best on that criterion. We call this task criterion-conditioned visual discrimination. VisionQ comprises (1) a corpus of 1,409 CVPR and ICCV papers with 1,800+ validated comparison figures and 3,911 hand-annotated data points linking method crops to author-stated visual claims; (2) a six-axis, 51-leaf taxonomy of the visual criteria behind qualitative judgment; (3) a criterion-conditioned evaluation protocol that hides method names, captions, and paper identity and reports accuracy per criterion; and (4) VisionQ-Judge, a DPO-tuned Gemma-4-E4B judge trained on symmetric evidence pairs, which reduces last-option predictions by 7.0pp and improves accuracy by 2.5pp on a held-out test set. Evaluating 20 open- and closed-source VLM judges, we find that the strongest reach only 63.1% accuracy (chance 32.2%) and that reliability varies sharply across criteria. Code: https://github.com/ReML-AI/visionq. Data: https://huggingface.co/datasets/visionq-anon-2026/VisionQ-1k.

Background: Protein language models (PLMs) are increasingly used for sequence generation and property prediction, but their size makes fine-tuning and deployment expensive. The effects of quantization and parameter efficient fine-tuning on performance, representations and generation remain insufficiently characterized. Results: We evaluated 4-bit quantization and low-rank adapter fine-tuning (QLoRA) across ESM-2, ESMC, ProtBERT, ProtT5, Ankh, Ankh3 and Profluent-E1. Across protein prediction tasks, many model-task pairs retained more than 90% of full fine-tuning performance. Peak GPU memory savings approached 90% for the largest models, although performance and efficiency varied by model, dataset and training configuration. QLoRA often preserved early-layer representations while inducing task-specific adaptations in middle and late layers, resembling full fine-tuning with smaller representational changes. Training speed and power effects were more varied. For unconditional generation with ProLLaMA, ProtGPT2, ProGen2, ProteinGLM and ESM3, 4-bit quantization largely preserved predicted structural and sequence-level properties, but token-level analysis revealed model-dependent shifts in autoregressive output distributions. Conclusion: QLoRA and 4-bit quantization reduce PLM computational requirements, particularly GPU memory usage. Our results support QLoRA as a first-pass strategy for memory limited adaptation, reserving full fine-tuning for challenging tasks, unstable architectures or low validation recovery. For generative PLMs, sequence-level and structural metrics should be complemented with distributional analysis, since downstream predictions alone may miss quantization-induced shifts. These approaches can broaden access to large-scale protein modelling while requiring model- and task-specific validation.

Bengali secondary education lacks curriculum-grounded benchmarks for STEM question-solving, and general-purpose language models struggle with the precise terminology, unit conventions, and derivations that physics problems demand. We introduce PhysicsMate, a benchmark of 1834 question-answer pairs built from the National Curriculum and Textbook Board (NCTB) Grade 9-10 physics syllabus and grounded in a multi-relational knowledge graph of 1760 nodes and 2600 edges across ten ontological types. We Low-Rank Adapt at 0.6B, 1.7B, and 4B parameters, with a unified recipe and demonstrate a significant increase in closed-book accuracy in all scales (+5.5, +15.0, and +23.3 percentage points). A node-type analysis shows that the most benefited by adaptation is the structured curricular knowledge, which consists of physical quantities and named laws, while the least benefited is the loosely specified entity-level knowledge. The 4B model has been adapted and quantized to a small offline binary that can be used for local inference in resource constrained environments and offers a viable path to curriculum aligned physics support in environments with limited connectivity and hardware.

Backdoored large language models (LLMs) can behave normally on benign inputs while producing attacker-specified outputs under hidden triggers. Existing defenses span four stages—prior-training, in-training, post-training, and inference-time—and share one of two underlying strategies: either suppress backdoor learning (by filtering poisoned data or interrupting its acquisition during optimization) or learn, then purify (by repairing model weights or gating inputs after a fully backdoored model has formed). We propose a third strategy, learn, but channel: allow backdoor formation during training but route it into a designated, quarantined component that can be disabled at deployment. To this end, we propose Quarantined Expert Shutdown QES, a computationally efficient containment strategy built in a regularization-steered MoE-like setting. Specifically, given a poisoned dataset, QES augments a Transformer-based language model with routed expert-specific LoRA branches and lightweight routers, and uses auxiliary routing objectives to attract trigger-conditioned behavior into a designated expert while preserving benign capability elsewhere. At deployment, mitigation reduces to a single constant-time operation: zeroing the quarantined expert's routing weight, without trigger screening or further updating model weights. Empirically, our methods reduce the attack success rate (ASR) from 100% to 0-10% on most settings across two tasks, three attacks, and four model families, while downstream utility is often preserved or only modestly affected. These results establish learn, but channel as a previously unexplored regime for backdoor containment in generative LLMs.

Deep neural network-based Voice Conversion (VC) and Text-to-Speech (TTS) models have rapidly advanced, enabling realistic voice cloning with minimal input data. Such capabilities raise serious concerns over unauthorized cloning of speaker identities and the associated privacy and security risks. Current imperceptible adversarial protection methods rely on quality control losses that are highly sensitive to hyperparameter tuning and computationally expensive due to lengthy optimization. To address these limitations, we propose a fast protection method that generates perceptually constrained perturbations in the frequency domain under a psychoacoustic masking-based constraint. Our approach strictly enforces perceptibility bounds during adversarial training, eliminating the need for iterative quality balancing and significantly reducing computational cost. Experiments on multiple state-of-the-art VC and TTS models show that STM achieves competitive or superior protection performance with substantially better perceptual quality and up to $45.3\times$ speedup over existing white-box baselines. These results demonstrate the effectiveness of frequency-domain perturbations with perceptual constraints as a practical paradigm for protecting against voice cloning.

Diffusion large language models (dLLMs) generate text by denoising a sequence or successive blocks, allowing several tokens to be revealed in parallel. Reinforcement learning with verifiable rewards (RLVR) reuses terminal feedback across these decisions, even as their conditioning context changes. We propose stepwise risk-sensitive GRPO (StepRS-GRPO), which varies the risk coefficient of the group-advantage transformation across denoising states while retaining the underlying trainer. For binary rewards, we show that this transformation is exactly a prompt- and state-dependent rescaling of centered outcome advantages. A capability-based calibration suggests a coefficient scale, while endpoint and interpolation ablations guide schedule selection. Across multiple dLLM backbones and mathematical reasoning benchmarks, StepRS-GRPO improves both pass@1 accuracy and pass@k coverage over centered GRPO, while increasing answer diversity. In our ablation studies, mass-matched controls support the contributions of state allocation and schedule direction, and the gains persist after matching the root mean square (RMS) of the advantages to that of centered GRPO. Reasoning-trace diagnostics further show that the diversity gains from StepRS-GRPO extend beyond final-answer strings.

We study fixed-time input-to-state stability (FxT-ISS) for discrete-time systems subject to exogenous inputs using finite-horizon Lyapunov functions. In the proposed Lyapunov- based characterization, the effect of the initial condition is eliminated after a uniformly bounded number of iterations. This characterization encompasses existing extinction-based Lyapunov conditions while allowing more general finite-memory mechanisms. Building on this framework, we establish a finite-horizon small-gain theorem for interconnected discrete-time systems, providing conditions under which FxT-ISS is preserved under feedback interconnections. The proposed framework is illustrated through examples arising in distributed algorithms, optimization, and feedback optimization.

Recent optical and near-IR photometric databases like Gaia data release 3 (DR3) and the VISTA Variables in the Vía Láctea eXtended (VVVX) ESO public survey allow the exploration of candidate unresolved multiple systems (binaries, triples, quadruples, etc.) in nearby clusters, that need spectroscopic confirmation. NGC 2437 (M 46) is a benchmark open cluster with well determined physical parameters. Its relative proximity (D=1.6 kpc) makes this cluster an ideal laboratory to measure the fractions of unresolved multiple systems. We aim here to identify unresolved triple and higher order systems that are reliable members of the open cluster NGC 2437, and provide the spectroscopic confirmation using radial velocities. We confirm that there is a sequence of unresolved binaries in NGC 2437, and report the discovery of a sequence of unresolved triple systems in this cluster. We corroborate membership for 13 binary systems and 8 triple systems using radial velocities measured by the ESO KMOS VVVX-GalCen spectroscopic survey. We further select stars with colors consistent with K-dwarfs, in order to have a controlled sample, comparable to the Solar neighborhood. We also find that the fraction of unresolved triples and quadruples is higher than that of the older open cluster Trumpler 19, studied using a similar setup, suggesting that these systems merge to form blue stragglers or get dissolved with time. We report the spectroscopic confirmation of 95% of the selected VVVX targets, thereby securing the presence of a sequence of unresolved triple systems in NGC 2437, located above the sequence of equal-mass binaries. The single stars constitute 60% of the mass of the cluster, while the rest of the mass (40%) is lodged in the unresolved multiple systems.

Long-form text-to-speech must retain requested words over extended generations, yet sentence-level training and evaluation can hide omissions and early stops. We introduce Balalaika-Longform, an open Russian corpus of 189 hours in continuous units of 30 seconds to 15 minutes. Long units and matched short windows support fine-tuning comparisons on the same source recordings, and the accompanying evaluation retains every synthesis attempt. We fine-tune CosyVoice3, Qwen3-TTS, VoxCPM2 and F5-TTS on either view and test continuous synthesis on 50 text-voice pairs with voices unseen in fine-tuning, at about 75, 300 and 1,200 words. At 1,200 words, CosyVoice3 WER falls from 99.9% after short-window fine-tuning to 47.3% with long targets, and to 16.6% with punctuated, format-matched transcripts; paired intervals favor long targets for Qwen3-TTS and VoxCPM2, and by a small margin for F5-TTS, whose absolute WER stays above 90%. The results isolate the effect of training sequence length under a fixed continuous-generation protocol.

We study fixed-time input-to-state stability (FxT-ISS) for discrete-time systems subject to exogenous inputs. We introduce Lyapunov-based sufficient conditions for FxT- ISS based on a class of dissipation functions with a fixed-time extinction property, which encompasses commonly used two- power constructions while allowing substantially more general decay mechanisms. Building on this framework, we establish a small-gain theorem for interconnected FxT-ISS systems in the presence of disturbances. The results are illustrated through several examples, including perturbed optimization algorithms and a discrete-time feedback-optimization architecture.

Observing what a large language model computes during autoregressive inference–online and without training probes–remains difficult. We introduce Lingtai, a training-free concept telemetry layer: at each generation step, residual states are projected onto a domain-specific bank of named concept anchors, constructed without labeled concept examples, outcome labels, gradient fitting, or activation-space optimization, producing a structured per-step concept-coordinate signal. Across code generation and grade-school mathematical reasoning, this signal exhibits a robust association with predictive uncertainty: the association survives problem-identity and token-position controls and is not attributable to a single token type, is not explained by a simple correct/incorrect mixture on GSM8K, and is not reproduced by matched random anchors; it is markedly weaker or direction-inconsistent in K-means and PCA projections. Two structures emerge: a recurring uncertainty-linked activity signal whose functional geometry is task-conditioned (distinct activity-entropy shapes on HumanEval, MBPP, and GSM8K), and an execution-specific trajectory identity with strong local inertia but weak re-instantiation invariance–under completion-only elastic alignment, corruption at k=32 (approximately a median quarter of the completion) on the matched re-execution subset still retrieves the archived episode at 62.0%, while a fresh execution retrieves it only 11.7-16.0% of the time. Finally, a matched audit finds no evidence that the scalar concept-activity signal used here supplies a stable correctness coordinate under the tested protocol; we therefore treat correctness as externally supplied. Telemetry adds 0.7-1.6% per-token decode overhead for the 161-anchor code implementation, with unchanged generated tokens.

We investigate fatigue failure in two-dimensional (2D) model glasses under cyclic shear deformation using atomistic simulations. We find that the number of cycles to failure diverges as a power law as the strain amplitude approaches the fatigue limit, with an exponent close to $-1$, in contrast to the exponent of $-2$ reported in three dimensions (3D). A failure exponent of $-1$ has also recently been observed in a 2D elastoplastic model, suggesting an interesting dependence on spatial dimensionality that needs to be rationalized. Measures of accumulated plastic activity, including dissipated work and non-affine displacements, exhibit scaling with the failure time that is consistent with results in 3D, and indicate a robust connection between damage accumulation and failure. To probe the origins of the variability of failure times, we perform isoconfigurational seeded simulations in which a localized soft region is introduced. While such seeding constrains the spatial location of failure, the distribution of failure times remains broad. These findings extend recent results in 3D concerning fatigue failure times to 2D glasses, including their relation to accumulated plasticity and apparent stochasticity, while also revealing a key difference, namely the exponent describing the divergence of the failure times.

Personalization has long relied on customer data to infer what an individual is likely to value. We call this customer evidence: the customer's historical behavior and preferences. Generative AI extends personalization by allowing providers to supply changing situational information at the moment a response is produced, without encoding every condition in advance. We define this provider-side context as information about what is possible, permitted, or advisable now. This flexibility creates a new problem: once context becomes easy to supply, more is not necessarily better. We develop a theory of context sufficiency in which the relevance of context to the customer's current intent matters more than its volume. The theory identifies four states, insufficiency, sufficiency, saturation, and interference, and introduces the Context-Sufficiency Frontier to locate the minimal relevant set. In a full-factorial experiment with a generative recommender at a large home-furnishing retailer, relevant context improved appropriateness, while irrelevant context reduced it and destabilized retrieval. The framework shifts personalization from supplying more context toward identifying what the current interaction actually requires and enforcing constraints throughout the service process.

In this paper, we construct explicit formulas for the heat kernel of the Hodge Laplacian restricted to an invariant subspace of $1$-forms on the real hyperbolic space $\mathbb{H}^{n}$. Using warped product decompositions, the problem is reduced to component operators on specific subspaces, yielding precise closed-form expressions for both odd and even dimensions. We further establish Millson-type recurrence relations that connect heat kernels across differing dimensions. Based on these explicit representations, we derive sharp uniform estimates and long-time asymptotic behavior, extending classical scalar heat kernel bounds to $1$-forms on real hyperbolic spaces.

We study ground states, criticality, and variational compactness for genuinely multipolar $L^p$-Hardy weights, $1<p<n$. A ground-state representation across the poles gives a complete criticality classification of the full potential and characterizes finite-energy attainment. Removing a geometric interaction term produces a variational problem governed by concentration at the poles and infinity. A strict concentration gap yields compactness, positive extremizers, and criticality of the optimal operator; for $p\ge2$, it is also necessary for attainment. In the subquadratic range, the optimal truncated coefficient is below one and attained throughout the full-potential critical set. For admissible parameters with $1<p<2$ and $(n-p)/p<s\le m(n-p)/p$, arbitrary fixed pole configurations exhibit negative directions, attained critical states, and coercive subcritical regimes outside this set along positive-weight paths towards the simplex vertices. A single-pole degeneration formula and continuity in the weights establish this coexistence. In the superquadratic range, an interaction threshold determines a non-attained infinity-concentration interval. Under strict separation from pole concentration, its endpoint is critical with infinite energy and a compact phase follows. We prove separation for every bipolar configuration and in several multipolar regimes, including a vertex neighborhood uniform in the exponent. Near parameters strictly inside the quadratic attained region, the optimal coefficient crosses one continuously through $p=2$. The proofs combine ground-state remainders, capacity, weighted compactness, and critical Hardy barriers.

Agent evaluations are increasingly used to compare LLMs and inform deployment decisions, yet ranks can reflect not only the model but also the effects of the evaluation conditions such as the scaffolds or tasks. This makes reliability claim-dependent: an evaluation that reliably ranks deployed systems may not reliably rank underlying models. We ask which conclusions current agent evaluations reliably support and what additional evaluation would improve them. We develop a Bayesian variance-decomposition framework for sparse, imbalanced agent leaderboards and apply it to 22 benchmarks from the Holistic Agent Leaderboard and Harbor Index. The framework separates signal, performance differences relevant to the intended claim, from noise, irrelevant variation that can still change rankings. We find: (1) Reliability depends on the measurement goal. Fixed model-scaffold systems are ranked reliably (0.935-0.994), while underlying-model reliability is substantially lower (0.148-0.841). (2) Scaffold choice can change conclusions. Inter-scaffold reliability measures whether scaffolds preserve model rankings, showing that scaffold effects vary substantially across evaluations. (3) More tasks cannot resolve all uncertainty. Even infinitely many similarly constructed tasks improve model-ranking reliability of a benchmark by at most 0.097 when uncertainty is dominated by limited scaffold coverage. (4) Pooling diverse benchmarks can improve cross-task rankings at lower cost. For rankings across diverse agentic tasks, pooling benchmarks raises projected reliability from 0.44 to 0.75 at the same task budget and can reduce projected cost by up to 83%. Evaluation design should follow the intended claim: identify what a score or ranking should mean, diagnose what limits its reliability, and spend evaluation budget on the sources of uncertainty that matter.

The performance of AI coding agents is highly dependent on their underlying coding rules. However, existing coding rules are typically hand-crafted and fixed, making the process labor-intensive and often suboptimal. In this work, we propose RuleEvolve, a self-evolving framework for coding rules. RuleEvolve maintains a pool of candidate coding rules and iteratively improves them. In each iteration, it employs an LLM-powered mutator module to generate variants from existing candidates, and then uses a judge module to evaluate these variants and update the pool with the best-performing ones. Extensive evaluations across two coding-agent frameworks, four backbone LLMs, and three benchmarks demonstrate that RuleEvolve outperforms both manual engineering and existing prompt optimization baselines in terms of functional correctness of the generated code, code length, and/or generation cost (e.g., tokens used).

Synthetic speech detection is critical for audio security, but performance can degrade when labeled data are scarce and evaluation conditions differ from training. This study examines quantum kernel methods and lightweight neural models for cross-corpus audio deepfake detection under limited training data. We compare a Quantum Support Vector Machine (QSVM), a classical support vector machine (SVM), and a multilayer perceptron (MLP), all trained on frozen wav2vec 2.0 embeddings using a strict budget of 200 training samples. To match the qubit budget of near-term quantum hardware, embeddings are reduced to four dimensions using principal component analysis, and all models use the same reduced features. Experiments on ASVspoof 2019, ASVspoof 5, the ADD 2023 Challenge, and the In-the-Wild dataset show that under severe domain shift from ASVspoof 2019 to ADD 2023, the MLP degrades to near-random performance, with an area under the curve of approximately 50% and an equal error rate of 50.0%. In contrast, the QSVM maintains meaningful discrimination, achieving an area under the curve of 76.0% and an equal error rate of 27.0%. This advantage is not consistent across transfer directions. When trained on ADD 2023, the QSVM falls below chance on two of three transfers, while the MLP performs better. These results suggest that quantum kernel methods can be competitive under severe cross-corpus shifts and strict low-resource constraints, but do not provide a consistent advantage under near-domain transfer. We interpret these findings as an empirical characterization of quantum kernel inductive bias under distribution shift, rather than evidence of quantum advantage, since the four-qubit kernel can be simulated exactly on classical hardware.

AI coding agents are ubiquitous in engineering workflows amongst industry and academia. Yet, despite their use in app coding, relatively less attention has been paid to their ability to execute on production incident response. This emerging field, termed agentic site-reliability-engineering (SRE) contains benchmarks limited by (1) unrealistic environments, typically toy repositories (2) non-standard framework implementations and (3) simple static verifiers. We introduce Incident-Arena, a human-built benchmark of 20 carefully selected tasks grounded in real-world deployed open source software. Each task deploys a production application to an ephemeral Kubernetes cluster, injecting a fault from the config layer through underlying images, and a sustained load profile given the task requirements. We also present a novel verification method, going beyond static checks to functional verifiers, holding systems level metrics stable, while ensuring repairs are done safely. Agent trials run an average of 2.81M tokens and 41 turns, going beyond existing benchmarks, demonstrating agentic long horizon reasoning. Across 20 tasks and 3 application substrates, frontier models score below 64.3%, with failures extending from diagnosis/localization errors, through incomplete repairs and unsafe regressions.

The representations that language models learn for concepts such as months, weekdays, and places display consistent geometric structure: circles and saddle-shaped "Pringle" manifolds. Recent work traced these structures to $\textit{translation symmetry}$ in word co-occurrence statistics, deriving the observed Fourier geometry when co-occurrence depends only on distance on an abelian lattice of concepts. We demonstrate that more general notions of symmetry lead to equally structured predictions. Considering symmetries defined by arbitrary finite groups, compact groups, and homogeneous spaces, we prove that whenever the co-occurrence statistics of a word family are invariant under a group $G$, the learned word embeddings consist of matrix elements of the irreducible representations (irreps) of $G$. Circles and Pringles arise when $G$ is cyclic, in which case the irreps are Fourier modes. We verify the irrep structure in three experimental settings. (i) The cyclic group $\mathbb{Z}_{12}$: for the months of the year we recover the known circular geometry. (ii) A dihedral group acting on the major and minor triads: we unify two classical observations – that transposition and chord inversion form a group ($T/I$) acting on chords (music theory), which $\textit{implies}$ that the well-known "circle of fifths" emerges in learned chord embeddings (machine learning). (iii) We explain and reproduce a recently discovered spherical representation of celestial objects in large language models (LLMs) as a spherical-harmonic embedding derived from our theory. Our results demonstrate that the geometry of learned representations is often a consequence of the statistical symmetry of underlying data.

We introduce a finite-dimensional state-space representation of bisexuality in Dirac notation, specify normalized amplitude weights, and analyze the limits of inference from one single observation. We extend the basis for discussing attraction across genders, then define a distinct state space for gender identity with an illustrative normalized time-dependent trajectory. The model is an analogy, not a physical theory or a measurement procedure for human sexuality.

The recent LUX-ZEPLIN (LZ) data shows a nuclear-recoil event around $248\,\mathrm{keV}$, motivating scenarios in which dark matter (DM) preferentially scatters at large recoil energies. We investigate an inelastic scattering interpretation in terms of a simple extension with a fermionic DM field interacting with the Standard Model through a single dimension-6 operator. A tiny soft symmetry breaking term facilitates an endothermic transition while naturally suppressing elastic scatterings. Without any fine tuning, a simultaneous explanation of the LZ data and the observed relic density can be effected for a DM mass in the $\sim (0.2,1)$ TeV range.

The Approximate polynomial satisfiability problem (APS), introduced by Guo, Saxena, and Sinhababu (CCC 2018), asks whether the zero vector lies in the Zariski closure of the image of a given polynomial map. Specifically, for a field $k$ with algebraic closure~$K$, the problem asks whether $\boldsymbol 0 \in\overline{\boldsymbol f(K^n)}$ for a polynomial map $\boldsymbol f=(f_1,\ldots,f_m)$ with $f_i\in k[X_1,\ldots,X_n]$. APS is a natural topological analogue of Hilbert's Nullstellensatz, namely the question of whether a given system of polynomial equations has a common zero. APS captures several problems in algebraic complexity, including border rank, hitting sets for border classes, and null-cone membership; it is known to be NP-hard and in PSPACE. We show that APS lies in the Counting Hierarchy (CH) over both the rationals and finite fields, substantially improving the known PSPACE upper bound. Our proof builds on a recent breakthrough due to Andrews, Garg, and Schost (FOCS 2026) on deciding Hilbert's Nullstellensatz in CH. As a corollary, our result improves the complexity of certifying hitting sets for border classes from PSPACE to CH. We also give a polynomial-time reduction of Hilbert's Nullstellensatz to APS, valid in any characteristic. In characteristic zero, we give a reduction of APS to the decision problem for the existential theory of real closed fields. Overall, our results place approximate polynomial satisfiability closer in complexity to exact polynomial feasibility and as a byproduct give improved complexity bounds for several problems arising in approximative complexity.

Users experience different balance challenges while standing, walking, and turning in virtual reality (VR), yet most locomotion techniques apply the same visual behavior regardless of movement state. We present a real-time movement-smoothing design framework in this paper that organizes visual adaptations according to movement-specific balance demands. The framework introduces three design strategies targeting standing, walking, and turning, illustrating how state-aware visual adaptations can support postural stability during locomotion. This paper focuses on the framework design and implementation, while a user study is planned as future work. Our work guides the development of future context-aware VR locomotion systems that better support safe and comfortable navigation.

Synthesizer programming is a challenging task, particularly in configuring modulation: the time-varying control of parameters such as pitch, filter cutoff, and oscillator level by modulators such as Low-Frequency Oscillators (LFOs). Prior work has shown that modulation can be reconstructed from the clean audio through curve matching, yet the recovered parameters are often untransferable to modern synthesizers. We present an interpretable destination-aware modulation and waveform recovery pipeline that mirrors how musicians often recreate sounds on a synthesizer. First, the system identifies the modulated destinations; then it recovers the LFO shape associated with each destination, and estimates the oscillator waveform to match the source timbre. The training is supported by a differentiable synthesizer that incorporates a noise oscillator, and more modulation options than previous work. We carefully train the models for optimal perceptual quality, with extensive exploration of perceptual losses and adversarial training. Through objective and subjective evaluation, we show that predicting the destinations correctly is essential, our model excels in the modulation focused inverse synthesis tasks, and our Gammatone loss and CQT discriminator significantly outperform the traditionally-used MSS loss in improving perceptual quality. We provide audio samples from the subjective test.

Classical resistive tearing theory assumes a prescribed current-sheet equilibrium, but a finite normal magnetic field can evolve the background on the tearing timescale. For a transversely pressure-balanced Harris profile of initial half-thickness $a$ and asymptotic reconnecting-field magnitude $B_0$, threaded by uniform $B_n$ ($\xi=B_n/B_0$), we derive an exact evolution in ideal magnetohydrodynamics: counter-propagating Alfvenic fronts broaden the neutral layer and generate a central sheet-parallel flow with shear at its flanks. Inner-layer estimates recover $\xi_{\mathrm{crit}}\sim S^{-3/4}$ for breakdown of the classical resistive balance, where $S$ is the Lundquist number based on $a$. Numerical frozen-profile calculations show loss of central resistive dominance and decreasing tearing growth as the neutral layer expands. Eigenfunction comparisons at fixed wavenumber and predominantly Doppler-shifted propagation support a normal-field-modified continuation of classical tearing. The normal field supplies no net perturbation energy. Energy extraction from the reconnecting-field gradient weakens as the sheet broadens. The velocity-gradient contribution becomes comparable when the neutral-layer width approaches the initial sheet thickness, without reversing the decline in growth. For $S=10^6$, $\mathrm{Pr}_m=0$, and $\xi=10^{-4}$, the maximum growth rate falls by a factor of about ten from its initial value at neutral-layer half-width $w\simeq1.59a$, where the growth and broadening times become equal. Only about 0.18 additional upper-envelope e-folds are estimated between this crossing and $w/a=2$, where the growth rate is about 25 times smaller than its initial value. These results support dynamical suppression of further linear amplification over the sampled evolution, even though the instantaneous spectrum remains unstable.

We investigate the optical and thermodynamic properties of quantum Oppenheimer-Snyder (qOS) spacetimes in both the black hole and naked singularity regimes. By analyzing test-particle motion and employing numerical ray tracing of thin accretion disk models, we identify their characteristic optical signatures. We find that increasing the quantum correction parameter reduces the apparent shadow size. While qOS black holes exhibit dark inner shadows, naked singularities produce central bright rings, driven by an infinite repulsive core barrier, yet still retain key optical features such as photon rings. Furthermore, using a semi-analytical approach, we obtain higher-order photon-ring impact and gap parameters that closely align with our numerical findings. Notably, the gap parameter intervals for qOS naked singularities deviate from the Schwarzschild case, providing a concrete criterion for observational tests. Finally, in the black-hole sector, the application of Geometrothermodynamics (GTD) identifies the specific-heat divergence at the Davies point. By linking photon-ring observables to GTD curvature scalars, we reveal a direct connection between qOS black hole shadows, thermodynamic stability, and microscopic interaction structures.

We prove in ZFC that $\mathfrak d=\omega_1$ implies $\mathfrak a=\omega_1$, settling a question of Roitman from the 1970s. Here $\mathfrak d$ is the least size of a family that eventually dominates every function in $\omega^\omega$, and $\mathfrak a$ is the least size of an infinite maximal almost disjoint (MAD) family of infinite subsets of $\omega$. Consequently, $\mathfrak a\leq\mathfrak d$ whenever the continuum $\mathfrak c=2^{\aleph_0}$ is at most $\omega_2$. This answers negatively Shelah's question whether $\mathfrak d<\mathfrak a$ is consistent with $\mathfrak c=\omega_2$, and shows that his model of $\mathfrak d=\omega_2<\mathfrak a=\mathfrak c=\omega_3$ has the least possible values of $\mathfrak d$, $\mathfrak a$ and $\mathfrak c$ among models of $\mathfrak d<\mathfrak a$. The proof also gives a preservation theorem: from any dominating family of size $\omega_1$, we construct a MAD family of size $\omega_1$ that remains maximal in every outer model in which the given family is still dominating. Thus, under $\mathfrak d=\omega_1$, some MAD family is indestructible by every $\omega^\omega$-bounding forcing; it can moreover be chosen almost strongly separable, hence Cohen-indestructible. These results give negative answers, under $\mathfrak d=\omega_1$, to two questions of Hrušák and to one of Brendle, Guzmán, Hrušák and Raghavan.

World action models improve robotic manipulation by conditioning actions on predicted futures, yet existing tactile variants largely inherit video-generation pipelines that reconstruct future tactile observations through iterative denoising. Such prediction can become unreliable under deployment drift: small changes in contact position or force may substantially alter tactile pixels even when the underlying contact evolution remains predictable. We introduce TacDyn-WAM, a heterogeneous visuo-tactile world action model that predicts implicit tactile dynamics rather than reconstructing future tactile observations. It learns TacRep, a dynamics-aware tactile target space trained through masked spatio-temporal prediction on tactile clips and regularized by relational structure distillation. A visual expert and an Implicit Tactile Dynamics Expert predict future visual and tactile representations in separate target spaces while interacting through joint attention; the tactile expert predicts future representations and their changes at multiple horizons in a single forward pass, and a read-only tactile memory supplies the current tactile state. On UniVTAC, TacDyn-WAM achieves an average success rate of 81.5% using only the provided demonstrations, reaching state-of-the-art-level performance and remaining competitive with models pretrained on large-scale visuo-tactile trajectories. Ablations confirm the benefits of both tactile pathways and TacRep over pixel-reconstruction and static alternatives. On five real-robot tasks, TacDyn-WAM reaches 71.0% average success, and modest-scale pretraining raises it to 85.0%, further validating our method.

We establish non-asymptotic sample complexity bounds for the least-squares estimation of vector autoregressive models for exponentially stable systems with heavy-tailed noise based on a single observed trajectory. By assuming i.i.d. noise, bounded noise covariance, and persistent excitation, we show that the estimation error is $\widetilde{\mathcal{O}}(r^{1/2}T^{-1/2+1/p})$ under bounded $p$th moment for $p > 2$, where $T$ is the number of samples, $r$ is the noise dimension, and $\widetilde{\mathcal{O}}(\cdot)$ hides logarithmic terms. We also introduce a unifying approach to sample complexity analysis applicable to broad classes of noise distributions and showcase this by deriving error bounds for sub-exponential and sub-Gaussian noise distributions. Finally, we specialize our analysis to autoregressive models with exogenous inputs and show that the dimension factor of the error bound is independent of the model order.