Science & Mathematics

Audio-language models remain unreliable at extracting basic musical properties from audio, motivating pipelines that place a music information retrieval (MIR) tool between the audio and the model. The tool reports measurements such as keys, chords, and structural similarity, and the model is expected to derive the answer. We test whether it can. Holding the musical information fixed, we vary derivation depth: the number of reasoning steps between the tool output and the answer. When the report states the conclusion, all five models score near 100%, and they follow it even when it is wrong. One step away, accuracy drops. On modulation, open models fall to 57-81%, although they answer that single step correctly when asked directly, and reach 0% at greater depths. Frontier models degrade less, but still degrade. Adding the audio does not recover the loss. The bottleneck in tool-augmented music analysis is not access to musical information but reasoning from it, so tool design must account for how much derivation it leaves to the model.

Amodal video segmentation asks for the complete mask of an object in every frame, including the parts hidden behind other objects. The best current methods generate these masks with fine-tuned video diffusion models, which need many denoising steps and an extra depth network, and give a different answer for every sample. Self-supervised video world models of the JEPA family are trained on a closely related task: predicting the representation of masked space-time regions from the visible ones. We therefore treat occlusion as masking. The tokens that the object's visible mask does not cover are dropped, the pre-trained V-JEPA 2.1 predictor fills them in, and a light head decodes the observed and predicted features into amodal masks. The resulting model, OccJEPA (Occlusion-as-masking JEPA), is deterministic and 33.5x faster than the diffusion-based state of the art, and more accurate on SAIL-VOS, on TAO-Amodal, including a held-out test split (+4.00 AP50) and visible masks tracked by SAM2, and on Kubric. Controlled experiments trace the gain to the predictor's pre-trained weights rather than to its capacity, for world models from 87M to 1.85B parameters. Code is available at https://github.com/zgp110/occjepa.

Time series anomalies take diverse forms, and individual detectors encode different anomaly scoring principles that rarely cover all anomaly types. Recent unsupervised time-series anomaly detectors provide complementary scoring signals, yet deploying them as an ensemble preserves the inference and maintenance cost of every individual detector. We study whether heterogeneous teacher detectors can be distilled into a single lightweight detector in the unsupervised setting, where no anomaly labels are available to define a shared target. AnoMix addresses this by extracting weak pairwise supervision from the training-set anomaly scores of frozen teachers. It normalizes each teacher score sequence independently and trains the student with CoRank, a multi-teacher ranking-distillation objective that converts teacher score gaps into pairwise supervision through teacher consensus. Across seven benchmark datasets, AnoMix successfully integrates complementary teacher scoring principles into a single deployable model, improving over individual detectors and static score ensembles while using a compact student with under 80k parameters on the benchmark settings. A zero-shot variant further shows that the distilled ranking signal transfers to unseen datasets and can outperform large pretrained time-series foundation models for anomaly detection.

We show that for all polynomials $T$ and $b$, every relation in NTIME($T$) has a binary PCP with $O(1)$ non-adaptive queries, perfect completeness, soundness error 1/2 and length $(T+b)\cdot(\log(T+b))^{O(1)}$, which is perfect zero knowledge against every adaptive, possibly malicious verifier making at most $b$ queries. This improves the length $\tilde O(T+b^2)$ of Zeilberger, O'Connor, Gur and Spooner (2026), who observed that Dinur's gap amplification preserves zero knowledge. The simulator runs in expected polynomial time (or in strict polynomial time, with a failure symbol or with negligible statistical error). For $b\le T$ the length is $T\cdot\mathrm{polylog}\,T$, the form of the shortest known constant-query PCPs without zero knowledge. Unless NP $\subseteq$ BPP, the dependence on $b$ is optimal up to the polylogarithmic factor for polynomial-time simulators: zero knowledge is free up to linear query bounds. Zero-knowledge interactive oracle proofs achieved length $\tilde O(T+b)$ in 2016-17; we obtain it for PCPs. Gur, O'Connor and Spooner (STOC 2025) asked whether NP has such PCPs of quasilinear length (their Open Problem 1). This paper answers it for every polynomial query bound, with the simulators above in place of a strict polynomial-time perfect simulator. The proofs were developed with extensive AI assistance and have been checked by the authors. We mask the Reed-Solomon proof of proximity of Ben-Sasson and Sudan at every node of its recursion. By a locator lemma, at every node, whatever $k$ of its proof symbols reveal about its input is determined by $O(k)$ input values. We prove it with the coprime-order argument of Bafna and Vyas (2026); our treatment of domain points is new. Dinur's gap amplification gives constant query complexity. The query count and the polylogarithmic exponent are absolute but very large.

A selective classifier may be claimed to satisfy a risk-coverage contract if some rule from a predeclared menu (a threshold or a union of $K$ score intervals) attains coverage at least $κ$ and error at most $α$. Refuting this existential claim requires ruling out every permitted rule. The intersection-union test (IUT) achieves validity without a multiplicity correction, but its sample cost can increase with menu richness. We characterize this cost and compare it with the optimal audit. For a fixed interior risk target $α$ and gap $γ$, the minimax sample size for K-interval menus at coverage floor $κ$ is $K(1-κ)/(κγ^2)$, up to logarithmic factors, in the stated small-gap, large-menu regime; exact IUT attains this rate. The lower bound holds even with a known score distribution, while the upper bound extends to menus of VC dimension $d$. Procedures that refuse to reject whenever any rule appears feasible on the audit sample can pay a linear $K$ cost even when a square-root rate is possible. Reference-free testing reaches a smaller scale under constant-risk alternatives, whereas unknown heterogeneous risks can make the linear cost unavoidable. In simulations with $K=4$, the raw menu cost varies 7.7-fold between $κ=.5$ and $κ=.95$; after theorem-motivated normalization the endpoint values differ by approximately 5%. Held-out experiments with three VLM judges show positive exact-hiding windows in all 60 split-specific confidence-gate families, with 28/30 contracts chosen on audit-A remaining hidden on untouched audit-B. On a cleaner multi-annotator target, hiding is weaker and transfers in 3/10 resplits. Frozen image and multimodal classifiers show boundary behavior consistent with the IUT guarantee and strong power relative to practical certificates.

This paper proposes Role-Enhanced Zero-Shot Chain-of-Thought (RE-ZS-CoT), a prompting framework that systematically integrates domain-specific expert roles to enhance reasoning in large language models (LLMs) without task-specific examples. We assess two versions of our approach. In the first version, static role selection, we evaluate eight predefined expert roles across arithmetic, commonsense, and reading comprehension benchmarks to determine their impact on reasoning performance. In the second version, dynamic role selection, each model autonomously selects the single most suitable role for each benchmark, enabling adaptive task-role alignment without manual specification. Across 20 LLMs and seven benchmarks, RE-ZS-CoT frequently surpasses zero-shot and zero-shot-CoT baselines, achieving performance improvements of up to 44 percentage points. The results reveal a strong correspondence between static and dynamic versions; the benchmarks where static roles yielded improvements were almost always those where dynamic roles also outperformed baselines, though static selection achieved peak performance. Ablation studies further show that unconventional roles degrade performance, underscoring the necessity of task-role alignment. The framework's scalability is validated across model sizes (2b-14b parameters), with smaller models like Granite3.3-8b achieving remarkable proficiency in arithmetic reasoning. RE-ZS-CoT advances zero-shot prompting by unifying role specialization with chain-of-thought reasoning, offering a practical, resource-efficient solution for improving LLM capabilities across diverse domains.

We aim to address the problem of Out-of-Distribution (OOD) Mode Averaging, a critical problem that emerges when distilling Flow Matching (FM) based behavior policies in offline Reinforcement Learning (RL): When learning over multi-modal datasets, standard distillation inadvertently forces the policy to interpolate across different behavioral modes, causing the generated actions to collapse into invalid, OOD regions. To resolve this, we propose Value-Rectified Distillation (VRD), a novel framework that reformulates offline policy distillation as a value-guided generative trajectory alignment problem. By employing a dynamic coupling mechanism, VRD actively rectifies flow dynamics to explicitly decouple distinct behavioral modes, cleanly isolating high-value actions. Specifically, we instantiate VRD through two complementary algorithms: VRD-Q, which utilizes value-quantile behavior distillation to statistically filter out low-value behaviors, and VRD-R, which reorganizes the latent space to geometrically isolate high-value generative trajectories from the FM-based behavior policy. Extensive empirical evaluations demonstrate that VRD improves existing generative baselines across diverse continuous control and planning tasks on the D4RL, OGBench, and Minari benchmarks, especially over those multi-modal, non-expert datasets.

Hexagonal boron nitride (hBN) encapsulation has transformed the optical spectroscopy of semiconducting transition-metal dichalcogenide monolayers (MLs) by suppressing extrinsic disorder and enabling the resolution of an increasingly complex hierarchy of excitonic states. This Review provides a spectroscopy-oriented framework for understanding and identifying excitonic complexes in hBN-encapsulated Mo- and W-based MLs and their alloys. The discussion covers bright and dark excitons, Rydberg states, charged excitons, biexcitons, phonon-assisted replicas, and higher-order Fermi-sea many-body states, including hexcitons and oxcitons, with an emphasis on the experimental signatures that distinguish them. Particular attention is given to the combined use of photoluminescence and absorption-related spectroscopy, electrostatic gating, polarization analysis, excitation-power and temperature dependences, magnetic fields, and time-resolved measurements to establish microscopic assignments. The role of alloying as an additional degree of freedom for engineering the excitonic hierarchy, spin-valley configuration, and magneto-optical response is also highlighted. By organizing the rapidly expanding phenomenology around experimentally accessible fingerprints rather than individual material systems, this Review aims to provide a practical and broadly applicable guide for identifying, comparing, and engineering excitonic states in high-quality two-dimensional semiconductors.

Reward guidance methods sample an exponential tilt of a base density using a diffusion model trained with only samples from the base distribution. Existing methods use what we call external guidance: additive terms in the drift or multiplicative terms on the transition probabilities. We propose internal guidance which leaves the output of the base denoiser unmodified and instead changes the point at which it is evaluated. For discrete distributions embedded in the hypercube $\{-1,1\}^{d}$, the denoiser of any linearly tilted density is exactly the base denoiser evaluated at a shifted and rescaled input. In practice the base denoiser is a neural network trained on noised samples of the base distribution. The required evaluation points for sampling the tilted density would be out-of-distribution for basic noising of the base distribution. We modify the training algorithm with random diagonal rescalings of the data, so that the denoiser network learns both in- and out-of-distribution evaluations of the denoiser relative to the base distribution. A single denoiser network samples binary reward tilts, class labels, and infilling with no fine-tuning and no external guidance term.

World models predict how embodied systems evolve under actions, but their accuracy can degrade when physical conditions move outside the training distribution. We study two components of causal modeling: which inputs a predictor may use, and how its mechanism computes their effects. In a synthetic benchmark with shortcut features, a frozen dependency mask derived from paired interventions reduces mean out-of-distribution (OOD) RMSE from 1.769 for a fully connected predictor to 0.697, close to the generator-oracle mask (0.694). For robot link-mass extrapolation in CausalWorld, with disjoint training and test mass ranges, we develop finger-local world models in which link masses enter inertia and dynamics-bias terms through explicit affine coefficients, actions act through either a learned torque network or the known PD controller, and learned residuals capture contact-related effects. Compared with a separately trained control whose mass inputs are fixed at training means, the known-PD model turns out-of-range masses into a consistent predictive gain: median time-averaged joint R2 over 150-step rollouts rises from 0.609 to 0.727, with improvement in all ten task-seed runs (exact two-sided sign test, p = 0.002). In a shared 100-step forecast, the learned-torque model is competitive with the strongest recurrent robotic world model (RWM) configuration (median joint R2 of 0.755 - 0.768 versus 0.752), whereas supplying out-of-range masses sharply degrades the MLP and recurrent predictors. These results indicate that how a physical parameter enters the dynamics mechanism determines whether its out-of-range value helps or harms prediction.

Abdominal organ segmentation is essential in modern medical applications. While existing deep learning approaches demonstrate promising performance, their heavy reliance on large-scale image data raises concerns over generalization capability and privacy risks. This motivates the development of generalizable segmentation training paradigms based on domain-randomized image synthesis, facilitating generalization across unseen real image domains. However, existing paradigms either suffer from limited segmentation performance when applied to abdominal organs, or require dataset-specific finetuning. To address these limitations, we propose ReSeg, a novel two-stage retraining-free framework for domain-agnostic zero-shot abdominal organ segmentation. At the first stage, our organ-agnostic synthetic data-driven pretraining approach exploits semantically and visually diverse synthetic images generated by a dedicated data engine to optimize a Generalizable Semantic Representation Network (GSRNet). The network is thereby capable of learning semantic features robust against variations in anatomical structure, imaging contrast, resolution and spatial context, which provide universal representation capabilities that serve as a foundation for segmentation learning. Subsequently, in the second stage, we continue to train an Organ-specific Encoder-decoder Segmentation Network (OESNet), whose encoder is initialized with the pretrained weights. This training process is facilitated by another dedicated data engine, which generates diversified visual images while strictly retaining the inherent spatial layout of abdominal organs. Comprehensive evaluation on seven abdominal organ segmentation datasets across three different modalities including CT, MRI, and PET demonstrates that ReSeg consistently outperforms existing state-of-the-art (SOTA) methods while eliminating the need for real training images.

Objective.To distinguish the contributions of acquisition context and photoplethysmography (PPG) to session-calibrated blood pressure (BP) tracking.Approach. The original comparison used three repeats of five-fold participant-level cross-validation in 653 AuroraBP participants, excluding calibration-labelled prediction targets. Matched analyses separated context and PPG and removed elapsed-time terms. Later exploratory follow-ups retained PPG, accelerometer signals and three initial cuff readings. Separate studies evaluated pulse-rate fusion and an intraoperative VitalDB model. Main results.The original wearable-state model achieved participant-level mean absolute errors (MAEs) of 7.697/6.139 mmHg for systolic/diastolic BP (SBP/DBP). Added context reduced MAE by 0.436/0.126 mmHg (97.5% intervals 0.305-0.573/0.053-0.200). In the matched factorial without the external PPG prior, PPG improved on calibration, time and protocol-provided posture/activity by 0.214/0.135 mmHg; removing time left the DBP interval including zero. Standard PPG descriptors contributed, but additional nonlinear/attractor benefit was not established. In a separate follow-up series, a sensor-history ensemble achieved 7.299/5.445 mmHg versus its static-rich reference's 7.363/5.470, small gains of 0.064/0.025 mmHg. Larger DBP errors did not improve uniformly. Quality-weighted fusion reduced pulse-rate MAE from 3.985 to 3.690 beats per minute. VitalDB MAEs remained 15.375/8.833 mmHg despite small PPG gains. Significance. The contribution is a calibrated evaluation framework with explicit controls, reproducible feature-cache-to-results analyses and retained negative findings. The sensor-history gain is exploratory after repeated dataset reuse. Independently estimated wearable states, complete raw-signal reproduction of the supplied richer features, external transfer and clinical adequacy remain unestablished.

Multilingual large language models are increasingly deployed in settings where visible reasoning traces support verification, education, and safety auditing, yet standard evaluations typically measure only final-answer agreement or holistic response similarity. As a result, they establish outcome consistency without determining whether users across languages receive comparable supporting evidence. We introduce a trace-level divergence framework that compares visible reasoning traces as structured sequences, decomposing agreement into structural similarity, order consistency, and holistic similarity, with pair-specific calibration to enable comparison across languages. Across three reasoning benchmarks, seven languages, and six frontier LLMs, the natural baseline---embedding each whole response and comparing the two by cosine---systematically overstates agreement: $52.1\%$--$67.3\%$ of examples in the top quartile of holistic similarity are not in the top quartile of structural similarity. This dissociation is stable across sentence encoders and step segmentations, and blind bilingual annotation studies in two English-anchored language pairs show that human judgments align with structural similarity rather than with the holistic baseline. A same-language control shows that the level of divergence is language-specific: two English traces for paraphrases of the same question agree more closely than an English trace and its counterparts in other languages. Our results show that answer-level and holistic evaluations can substantially overestimate the consistency of multilingual explanatory artifacts, motivating trace-level evaluation for reliable multilingual deployment.

Understanding how vision-language models use visual information requires connecting internal representations to model behavior. We present vlmprobe, an open-source library that integrates lens readouts, attention attribution, causal interventions, and sparse autoencoder (SAE) analysis within a common model interface. Shared adapters, calibration and checkpoint utilities, evaluation pipelines, and visualizations support systematic analysis across architectures. Using this framework, we compare methods across vision-language, language, and vision models through matched case studies and controlled evaluations of pretrained sparse dictionaries. The comparisons show that representation readouts and spatial explanations depend on the analysis objective and observation site, and that local attribution estimates can disagree with measured intervention effects. SAE evaluations further expose distinct trade-offs in reconstruction fidelity, concept readout, and spatial localization. Together, these results support combining multiple forms of evidence to evaluate an interpretation. vlmprobe provides a reusable basis for selecting methods, comparing their conclusions, and testing hypotheses about model behavior. The code is available on github.com/ubc-tea/vlm-probe

We introduce MixAgent, an agentic pipeline for music post-production that follows the workflow of audio engineers to build interpretable and editable mixing plans guided by a reference song. A grouping agent first routes related input tracks, such as drum recordings, into subgroups. A planning agent then uses retrieved examples to choose and order audio effects, such as volume control and reverb, for tracks, subgroups, and the full mix. To refine the mix, a detector identifies what needs correction, and a critic proposes numerical parameter adjustments from tracks to the full mix. In our experiments, prompting recent multimodal models to listen and revise often worsens the mix, making reliable correction a central challenge. We therefore build MixCritic, a dataset of 45,657 repair critiques derived from parameters fitted to professional mixes, and train the critic with supervised fine-tuning followed by preference optimization on rendered audio outcomes. On MedleyDB, learned refinement reduces error by 19% relative to the initial mix, with preference optimization reducing it by 8.4% over supervised training alone. Retrieved examples further improve effects selection, and a human listening test shows that MixAgent gets higher quality ratings than mixes produced by Qwen3-Omni, Gemini 3.8 Flash and GPT-4o-audio.