Papers

Embodied systems must make knowledge acquired during one encounter usable in another despite changes in viewpoint, motion, and illumination. Yet aggregate cross-video accuracy conflates failures of local perception with failures to preserve observation identity, establish correspondence, and compose evidence, obscuring whether local video understanding actually transfers. We introduce EgoGears, a complementary single- and multi-video benchmark designed to diagnose this transition. It contains 567 single-video and 1,487 multi-video questions derived from 126 human-collected egocentric recordings covering 39 outdoor routes. Repeated traversals across movement speeds and lighting conditions ground comparisons in shared physical environments; 531 questions require alignment across independent recordings. Single-video questions measure the local visual, spatial, and motion evidence available to a model, while multi-video questions test whether evidence remains bound to the correct observation and can be composed into consistent route relationships. We report 29 single-video and 31 multi-video MLLM configurations across six model families in the main leaderboard. Among the 20 configurations evaluated comparably on both splits, every model performs worse on multi-video questions, with a mean decrease of 22.5 percentage points, and the gap persists when answer format and scoring are held fixed. The gap is not explained simply by additional videos or recording boundaries. The central bottlenecks are observation–evidence binding and ordered route-state tracking. The code and benchmark are publicly available at https://github.com/lei-qi-233/EgoGears.

How much of an image does a detector need to see? Small RGB regions can retain useful evidence of image synthesis even when they reveal little of the full scene. Motivated by single-patch detection, we study patch-wise supervision: a shared backbone classifies explicit crops, each crop receives its own loss, and patch probabilities are averaged only at inference. The procedure requires neither handcrafted residual filtering nor a learned image-level fusion module. Experiments span single-patch selection, multiple generator collections, and four CNN and Transformer backbones. On GenImage, the reported patch-wise variants improve average accuracy over their whole-image counterparts across all four backbones. Comparisons of supervision granularity, source resolution, crop size, and inference coverage further characterize the approach, while post-processing tests and difficult-image evaluation reveal its limitations. The historical experiments include evaluation-based model selection, so their scores are not presented as a uniformly selected leaderboard comparison. Overall, the study identifies explicit local input and patch-level supervision as a simple, useful combination for investigating generalizable AI-generated image detection.

We compute the Dolbeault cohomology of minimally stable polytopal LVM manifolds using a finite curvature complex. Let $B$ be the indispensable weight block. A canonical exact sequence identifies the kernel of the curvature map with $(\operatorname{coker}B)^\vee$ and its cokernel with $(\ker B)^\vee$. At full rank, the Dolbeault groups are direct sums of reduced cohomology groups of induced subcomplexes of the visible simplicial sphere, tensored with the exterior algebra on the complex dual of the deck lattice. One proof uses the curvature model and Stanley–Reisner Tor; a second gives an additive comparison through completed Laurent expansions and holomorphic descent. Full curvature rank is equivalent to Frolicher degeneration at $E_1$. Every manifold in this class fails the $\partial\bar\partial$ lemma. For polygons, the number of facets and the curvature rank determine the entire Hodge diamond. We construct a proper holomorphic family over a disk with fixed visible square whose curvature rank drops at the origin. The total dimension of Dolbeault cohomology increases there by sixteen. At full rank, holomorphic descent also computes tangent-sheaf cohomology, including resonant contributions in positive Cech degree. We determine the Kodaira–Spencer map of the normalized weight family and give a cohomological criterion for semiuniversality with smooth base. For a resonant example over the square, the global vector fields have a polynomial basis of nine elements and $h^1(\Theta)=19$. The primary obstruction map is nonzero. The additional Cech class integrates in a holomorphic family.

Deep Support Vector Data Description (Deep SVDD) has become a prominent framework for unsupervised anomaly detection by learning latent representations that compactly characterize normal data around a center. Despite its empirical success, anomaly decisions produced by Deep SVDD are typically made solely based on anomaly scores without rigorous statistical guarantees, thereby limiting their reliability in safety-critical and high-stakes applications where false positives must be strictly controlled. In this paper, we propose PADI (Post-Anomaly Detection Inference), a novel framework that equips a trained and frozen Deep SVDD detector with statistically valid inference by leveraging the Selective Inference framework. Specifically, PADI performs inference conditional on the event that a test instance is identified as anomalous by Deep SVDD, thereby enabling rigorous statistical assessment of anomaly decisions. Based on this formulation, we derive valid selective p-values that quantify the statistical significance of the detected anomaly. Using these p-values, we theoretically establish control of the false positive rate (FPR) at a user-specified significance level $\alpha$ (e.g., $\alpha=0.05$). Furthermore, we extend the proposed framework to Deep Semi-Supervised Anomaly Detection (Deep SAD), providing a principled approach for statistically reliable inference in semi-supervised anomaly detection settings. Extensive experiments on both synthetic and real-world benchmark datasets robustly support the theoretical findings. The results demonstrate that PADI consistently achieves proper FPR control while attaining superior true positive rates compared with existing approaches.

Forecasting time-varying functional connectivity from electroencephalography (EEG) requires modeling both history-dependent trends and structured variability across channels. Conditional flow matching provides a framework for distributional forecasting, yet it remains unclear whether graph-informed source distributions offer practical advantages over isotropic noise and strong deterministic predictors. We introduce a graph-structured residual flow framework that separates conditional mean prediction from stochastic residual transport. A history-only predictor estimates the future connectivity graph, while a graph Gaussian source encodes dependencies derived from past connectivity through a Laplacian-based covariance. A conditional velocity field transports source samples to future graph residuals, with transport time explicitly distinguished from physical EEG time. Our study identifies the conditions and controls needed to distinguish useful residual transport from improvements attributable to deterministic prediction, learned representations, and sampling effects.

We study g-dimonoids, which are algebraic structures equipped with two associative binary operations satisfying a prescribed system of axioms. We develop a construction of g-dimonoids based on inflations of semigroups, construct several classes of 2-dinilpotent commutative g-dimonoids, and investigate the isomorphism problem for these classes. As an application of these results, we obtain a structural description of all commutative g-dimonoids of order 3, up to isomorphism.

Systems operating in dynamic environments require timely updates to sustain performance. For resource-intensive systems such as machine learning models and digital twins, strategically timing updates is essential. Updating too frequently wastes resources, while updating too infrequently leads to costly performance degradation. The problem is particularly challenging when the system's degradation pattern is unknown a priori, as is common in new operating environments. We formalize this challenge as a novel timing bandit problem, where each arm represents a candidate update interval with a fixed update cost and an unknown, stochastic degradation cost. Three structural properties distinguish this setting from standard multi-armed bandits: selecting an interval commits the learner to multiple time slots before the next update; arm costs are composed of per-step degradation costs and a fixed update cost; and selecting a longer interval naturally reveals degradation at every intermediate step, providing consecutive feedback relevant to shorter intervals. By exploiting these structures, we develop Balanced Consecutive Arm Elimination (BCAE). BCAE achieves $\tilde{O}(\sqrt{T})$ regret, improving upon the $\tilde{\Omega}(K\sqrt{T})$ regret of standard bandit algorithms in this setting, where $K$ is the number of candidate update intervals. We further propose an Optimism-Enhanced variant (OE-BCAE) that integrates lower-confidence-bound principles to improve empirical adaptivity while preserving the same regret order. Moreover, the regret bound achieved by our algorithms matches the theoretical lower bound up to logarithmic factors. Simulation results demonstrate that our algorithms achieve low regret and remain stable as both the number of arms and the update cost vary.

We implement a Macaulay2 package for computations involving rings of truncated Witt vectors $W_n(R)$ of a finitely generated $\F_p$-algebra $R$. This package includes ring operations (addition, multiplication, Frobenius, Verschiebung, etc.) on elements of $W_n(R)$, conversion between tuple representatives and ghost map representatives of elements of truncated Witt vectors over polynomial rings, and explicit computation of $W_n(R)$ as a finite-type algebra over $\Z/p^n$. This functionality is based on an algorithm we developed for performing arithmetic operations in rings of finite length $p$-typical Witt vectors over finitely generated $\F_p$-algebras. In addition, we implement an algorithm to calculate lifts of Frobenius from $\F_p$-algebras to flat lifts over $\Z/p^2\Z$, and an algorithm to find the quasi-$F$-splitting height of a local or graded complete intersection ring.

Persistent textual memory allows language models to carry information across long interactions, but learning what to remember is fundamentally a credit-assignment problem. A memory rewrite may only become useful many steps later, while much of the observed utility may be inherited from information already stored before the rewrite. We introduce Memory Gain Policy Optimization (MGPO), which isolates the incremental value of each memory rewrite by crediting it for its marginal contribution to current and future downstream utility. This turns delayed memory utility into a direct learning signal for optimizing what information should persist. We study MGPO on document-level information extraction, where structured supervision makes the effects of individual memory updates directly measurable. MGPO improves extraction while reducing average memory length by nearly 80% relative to the initial memory policy before optimization. The learned memory policy also supports reuse and transfer across domains, downstream models without further training. These results show that effective memory learning depends not only on preserving useful information, but on identifying which memory updates create lasting incremental value.

Building on the bit-complexity framework of Raghavendra-Weitz and the moment-SOS criteria of Gribling-Polak-Slot, we study the effective use of truncated vanishing identities over Boolean polynomial systems. A complete, constructible identity space gives an augmented moment SDP that can be optimized with exact rational feasibility and arbitrary additive accuracy. We give a self-contained geometric implementation: an explicit affine reduction and a simplex of Boolean evaluations supply the radius bounds required by rational ellipsoids. The identity space can be constructed directly or extracted from a supplied graded Groebner basis, including one supplied at a higher truncation degree. An explicit transfer theorem connects this augmented formulation to the original system. Two-sided SoS derivations eliminate the added equality axioms from certificates, with controlled degree and coefficient growth, and imply containment of a projected higher-level moment relaxation in the augmented body. Together with spectral coefficient bounds, this gives polynomial-time search for rational proofs with an additive perturbation. For Min-closed linear systems, propagation constructs the identity space in polynomial time at fixed degree and gives degree-(4t+4) certificates for both signs of each degree-t basis element. Consequently, a rational degree-(8d+4) proof of f + epsilon >= 0 can be found in polynomial time for fixed d whenever f >= 0 has a degree-2d proof. The augmented degree-2d moment SDP can be optimized in polynomial time with exact rational feasibility and a comparison to the original degree-(8d+4) relaxation. Boolean complementation gives the same results for Max-closed systems, including generalized packing and covering. Min-closed systems thus provide a concrete application of the general criteria. All complexity bounds are in the Turing model.

X-ray pulse-profile modeling can measure neutron-star radii, but its physical interpretation remains limited by the largely phenomenological description of the surface hotspots. Here we connect the X-ray hotspots directly to the pulsar magnetospheric structure and jointly infer the mass, radius, and magnetic geometry of the massive millisecond pulsar PSR J0740+6620 from NICER and XMM-Newton observations. Using a GPU-accelerated implementation, we compare two low-dimensional magnetic configurations that can produce non-antipodal emission: a centered axisymmetric dipole–quadrupole and a shifted dipole. The shifted-dipole model is decisively favored under the adopted priors, with $\Delta\ln Z=5.65$ (Bayes factor $\sim285$), and yields $M=2.059^{+0.068}_{-0.069}\,M_{\odot}$ and $R=13.28^{+1.47}_{-1.07}\,\mathrm{km}$, broadly consistent with previous phenomenological analyses, while requiring nearly symmetric heating of the two polar caps. Both the marginalized shifted-dipole posterior and the timing-compatible representative require an effective X-ray-emitting footprint much smaller than the nominal force-free polar cap calibrated for a centered dipole. This compact footprint is not reproduced by our exploratory photon-conversion estimates and instead points to missing physics in the mapping from magnetospheric currents to observable X-ray emission. Our results show that physically motivated hotspots can preserve the mass–radius inference while turning hotspot morphology into a probe of pulsar magnetospheric structure and emission physics.

A key objective in Weyl-semimetal (WSM) research is to identify phases built from minimal configurations of Weyl points (WPs), which provide the simplest settings for investigating the intrinsic physics of chiral topological charges. Pang et al. [Phys. Rev. Res. 8, 033304 (2026)] recently established that crystallographically realizable WSMs can be decomposed into linear combinations of 16 minimal "irreducible Weyl molecules" (IWMs). A fundamental question remains unresolved: which of these non-decomposable configurations can themselves form standalone crystalline phases? Here we introduce irreducible Weyl semimetals (IWSMs), in which two bands crossing near the Fermi level form an elementary symmetry-compatible complex whose complete Weyl-node configuration is charge neutral and crystallographically indivisible. Combining all 1651 magnetic space groups with Weyl-orbit multiplicities and band-compatibility relations, we classify IWSMs formed by crystalline-symmetry-protected twofold WPs. Only ten of the sixteen primitive inventories admit one-copy realizations, comprising four Pair, four Split, and two Mixed charge-node architectures. Remarkably, this classification uncovers five unconventional and previously unrecognized topological phases: the charge-three Pair IWSM {3, -3}; the Split IWSMs {3, -1, -1, -1}, and {4, -1, -1, -1, -1}; and the Mixed IWSMs {3, 1, -2, -2} and {3, 3, -2, -2, -2}. For all ten classes, symmetry-constrained lattice models verify the complete node inventories, Chern charges, and surface chiral-flow incidence. Our results establish crystalline irreducibility as a phase-level organizing principle beyond individual Weyl nodes and provide a symmetry-resolved framework for identifying minimal Weyl complexes in electronic, phononic, and photonic systems.

We determine the quantum query complexity of testing $k$-juntas for Boolean functions, unitaries, and quantum channels. Let $n$ denote the number of input bits or qubits and $\varepsilon$ the distance parameter. For all sufficiently large $k$, $n \ge 2k$, and $2^{-k/16} \le \varepsilon \le 1/16$, we establish the following tight bounds. $\bullet$ Junta functions: The query complexity is $\Theta(\sqrt{k/\varepsilon})$, closing the gap left by Ambainis, Belovs, Regev, and de Wolf (SODA 2016) and Bun, Kothari, and Thaler (STOC 2018, ToC 2020). $\bullet$ Junta unitaries: The query complexity is $\Theta(\sqrt{k}/\varepsilon)$, closing the gap left by Chen, Nadimpalli, and Yuen (SODA 2023). With forward-only access, the query complexity is $\Theta(k/(\varepsilon^2 \log k))$. At constant $\varepsilon$, these bounds establish a separation between $\Theta(\sqrt{k})$ queries with forward and inverse access and $\Theta(k/\log k)$ queries with forward-only access. $\bullet$ Junta channels: The query complexity is $\Theta(k/(\varepsilon^2 \log k))$, closing the gap left by Bao and Yao (COLT 2023, TPAMI 2025). The upper bounds for channels and forward-only unitaries are achieved by nonadaptive, ancilla-free testers using product inputs and single-qubit measurements. Our upper bounds follow from a reduction to union size testing, a generalization of support size testing to distributions over subsets of a ground set. We develop optimal testers for this problem both from independent samples and from a reversible sampler. Our matching lower bounds use moment-matching constructions and a phase-masked unitary ensemble. For this ensemble, the average acceptance probability of every forward-only quantum algorithm admits an exact representation in terms of independent classical samples.

Autoregressive (AR) video diffusion enables low-latency, streamable video generation, but prediction errors often accumulate over long rollouts. Training the generator on its own rollouts exposes it to these imperfect histories. However, existing video-level distribution matching distillation (DMD) scores the whole rollout jointly. Because a chunk is evaluated together with its past and future, its correction can favor matching artifacts in the surrounding context merely to preserve temporal consistency. To provide a clearer visual-quality signal, we introduce Rollout-Marginal Distillation (RMD). RMD retains the generated history for AR prediction but scores each chunk independently against a chunk teacher, ensuring its quality correction is not compromised by an imperfect temporal context. To compensate for the lack of temporal context in independent chunk scoring, RMD subsequently applies video-level DMD to restore temporal coherence. Extensive experiments demonstrate that RMD maintains high visual quality far beyond its training horizon and outperforms video-level DMD baselines. Code and video results are available at https://cjeen.github.io/RMD

Recent work on anchored diffusion language models improves denoising by shaping an intermediate latent space with supervised important-token targets. In this work, we introduce time-based (self-supervised) anchoring, which learns and reuses latent anchors without requiring such targets. Our key observation is that anchors encode persistent properties of the clean sequence, such as its semantic intent, global structure, or intermediate plan. Although their hidden representations become stale as the token canvas evolves, their semantic content remains useful across nearby diffusion times. This is implemented through a two-stage architecture consisting of a relatively expensive anchor network that generates the latent cache state and a lightweight denoising network that intelligently combines the cached latent state with the current state at each reverse step using a fusion module. This gives anchoring a latent-space caching interpretation: the anchor network is evaluated periodically, while its cached representation is reused across multiple reverse steps. We instantiate this framework as TADM:Post-train, which time-anchorizes pretrained DLMs, and TADM:Pretraining, which learns time-based anchors during pretraining. Applied to DiffusionGemma-26B, TADM:Post-train improves throughput by approximately 49% to 79% on several math, code, and STEM benchmarks (GSM8K, AIME26, GPQA-Diamond, LiveCodeBench-v6, HumanEval, MMLU-Pro). TADM:Pretraining reduces Transformer-layer computation by up to 38% relative to a standard single-stage DLM, achieves up to 73% higher measured throughput than ADLM.

Agents can learn from past executions, but enabling different agents to reuse and build on one another's experience remains challenging. We introduce EpiCon, a shared multimodal memory framework for agent collective learning without updating host model parameters. EpiCon links question-level memory evolution to a persistent experience bank through two independently trained 2B models: a memory controller and a tree self-organizer. The controller jointly refines textual guidance and visual evidence across attempts and selectively includes visual memory. The self-organizer consolidates lessons hierarchically and retrieves experience and rules for new problems. We evaluate EpiCon on eleven benchmarks spanning four multimodal task domains, using two harnesses and multiple backbones. A frozen bank improves other systems even with a single solving attempt. A second harness raises the original system's macro-average score by 2.6 points across eleven benchmarks. Across four host configurations, EpiCon improves macro-average scores by 1.7 to 4.9 points over No Memory and reduces memory-operation time by 67% to 74% relative to backbone-sized memory models.

Visual Language Action (VLA) models offer unprecedented generalization for autonomous robots; however, their real-world deployment is frequently bottlenecked by unreliable execution and the prohibitive computational cost of fine-tuning for specific robot embodiments and tasks. To bridge this gap, we propose WayFinder, an end-to-end, closed-loop hierarchical VLA framework that circumvents the need for fine-tuning by decoupling high-level task reasoning from low-level kinematic control. WayFinder utilizes a zero-shot, offboard Multimodal Large Language Model (MLLM) policy to process linguistic context and state maps for strategic waypoint generation. Asynchronously, a lightweight, onboard policy executes real-time kinematic control at high frequency based on continuous sensor feedback. We evaluate WayFinder in Microsoft AirSim, testing on four environments of varying complexity and three MLLM scales to balance prediction efficacy with computational efficiency. Our results demonstrate that WayFinder achieves superior navigation reliability compared to baseline low-level policies. By querying the high-level MLLM only during navigation failures, WayFinder eliminates the need for fine-tuning, minimizes expensive inferences, and significantly increases navigation success rates by up to 27.45%.

What are the inductive biases of a Transformer architecture? Existing theory on how the forward pass shapes representations either considers whether Transformers escape from rank collapse or demonstrates that self-attention drives tokens toward cluster patterns. The latter view arises from an elegant dynamical systems perspective, but relies on simplified architectural assumptions, and does not explain the rich structures observed in practice. This leaves a major open question: when a full Transformer escapes rank collapse, how does it structure token representations? Using pattern-formation theory, we show that the dynamical view of Transformers can account for Positional Encoding, Multi-Head Attention, and Output-Value geometry. We demonstrate that a full Transformer architecture imposes an inductive prior by selectively amplifying a rich set of previously unreported patterns, including traveling or rotating waves among others. We characterize the role of each architectural component in controlling which pattern is amplified, which ones stabilize, compete, or coexist. Finally, we show that these structures can act as a controllable dynamical prior that facilitates learning. By choosing both task-aligned positional encoding and weight initialization, we demonstrate improved data efficiency and accelerated optimization on controlled sequence tasks and with ConViT on CIFAR-10.

In 2018, Andrews defined and studied the partition function $\mathcal{EO}(n)$, which counts the number of partitions of $n$ where each even part is less than each odd part. Thereafter, Uncu considered a different subset of such partitions, namely those partitions counted by $\mathcal{EO}(n)$ where there are no repeated even parts (we denote the number of such partitions by $\mathcal{EO}_u(n)$). In this paper, we prove several congruences modulo powers of $2$ for $\mathcal{EO}_u(n)$. We also prove some infinite families of congruences as well as a recurrence for the number of this class of partitions. Our methods involve elementary, algorithmic, and modular form techniques.

We consider one-loop QCD corrections to Higgs-boson production via vector-boson fusion at hadron colliders. As an input for future exact NNLO computations, we derive the one-loop helicity amplitudes to higher orders in the dimensional regulator. We present a direct calculation of the amplitudes with spinor-helicity techniques and different $\gamma_5$ schemes. In addition, we compare this approach to a projector-based calculation. In order to derive analytical results to higher orders in $\epsilon$, we calculate the pentagon master integrals by the method of $\epsilon$-factorised differential equations and express them in terms of an algebraically independent set of basis functions.

Reasoning across videos requires aligning events, matching identities, comparing motion, and integrating partial observations. Evaluating these capabilities and testing how to improve them requires both reliable labels and targeted supervision. We introduce SYNCR, a simulator-grounded framework that connects these two needs through shared task generators. Built on Habitat, Kubric, and CLEVRER, SYNCR derives answers from environment state and provides 4,000 evaluation questions and 15,960 training questions over disjoint videos, spanning eight cross-video reasoning tasks. Visual ablations and human evaluation assess dependence on the supplied evidence and answer recoverability. Evaluation of 22 multimodal large language models reveals persistent difficulties in physical comparison and scene integration that increasing model size does not consistently resolve. Supervised fine-tuning raises Qwen3-VL-8B's average SYNCR accuracy from 32.6% to 61.6%, with gains extending to task configurations and video sources absent from training for those tasks. Transfer to real footage is most consistent for temporal ordering: accuracy improves by 9.0-20.5 percentage points on constructed Assembly101 and Panoptic ordering sets across three checkpoints spanning two model families and two model sizes, with additional gains on existing temporal reasoning benchmarks. These results establish SYNCR as a controlled setting for diagnosing cross-video reasoning failures, testing their learnability, and identifying where synthetic supervision transfers.

Projection pursuit searches for a direction along which the data look least Gaussian. When the observation space contains a large Gaussian complement, the empirical objective can be minimized by a direction that carries no signal, with empirical kurtosis as low as at the truth. Sample splitting exposes rather than repairs this failure. Appending coordinates independent of the latent regime degrades the search while leaving Bayes recoverability unchanged. Restricting the search to the column space of a known forward operator removes the failure exactly on the negative-kurtosis branch. Estimating a principal subspace from the data is the alternative. In a controlled two-component model, the leading sufficient scalings differ in the gain with which the operator transmits the discriminant: $\varsigma^{-4}$ for covariance-spike estimation and $\varsigma^{-8}$ for fourth-moment search. At fixed search dimension, the measured threshold ratio collapses onto $n/p^2$ with exponent $0.156$, close to the predicted $1/8$. This is an empirically supported scaling motivated by sufficient bounds, not a proved asymptotically tight law. When the search dimension is varied, the measured exponent is $0.325$, substantially larger than $1/8$, and the tested range does not identify its functional form. The crossing location also depends on calibration and model configuration. Under a downstream excess-error criterion, the scaling largely disappears.

Production machine learning (ML) stacks often split graph compilation and kernel execution across different layers and languages, making backend behavior, deployment guarantees, and performance fallbacks hard to reason about end-to-end. RLX addresses this gap with a single Rust codebase that combines compiler and runtime roles around one primitive-level, three-level intermediate representation (IR), plus a transparent dispatch contract that resolves each operator to native, common-IR, or rewritten lowering and fails compilation when legalization is not possible. The same IR targets fourteen runtime devices (cpu, metal, mlx, ane, cuda, rocm, oneapi, tpu, hexagon, gpu, vulkan, opengl, directx, webgpu) and two specialty codegen paths (Cortex-M INT8 and FPGA), ingests safetensors, GGUF, ONNX, and rten formats, supports F16/BF16/F64/C64 and quantized INT4/INT8 flows with AMP/PTQ/QAT, and scales via tensor-/pipeline-parallel collectives over TCP and RDMA transports. Beyond neural workloads, RLX also extends to scientific/physics-style domains through sparse and dense linear algebra extensions (e.g., CSR LU/CG/matvec and LAPACK- backed factorizations) and 3D Gaussian splatting operators. We evaluate RLX against PyTorch, TensorFlow, JAX, candle, burn, tch, rten, MLX, CoreML, IREE, Glow, TensorRT, and tinygrad under identical input generation and p50 measurement methodology on one host. On all-MiniLM-L6-v2, RLX-Metal is fastest at every batch (e.g., 16.6 ms at batch 32 vs. PyTorch-MPS 26.7 ms). In the MNIST training table, RLX also has the top-throughput entry (graph-fused MLP: 946,487 img/s), above NumPy+BLAS (787,349 img/s), while retaining 100% top-1 parity on reference checks (e.g., Qwen3).

On-policy self-distillation (OPSD) trains a student to match a privileged teacher distribution along its own sampled trajectory. Standard OPSD applies this supervision to unverified student rollouts while conditioning the teacher on privileged context, typically a reference solution. We separate these roles in a factorial analysis and find that scaffold correctness has a stronger effect on downstream accuracy than context correctness. Unverified scaffolds create an imitation gap because the teacher can use information unavailable to the student. This gap shrinks with model scale, yet OPSD continues to supervise mostly unverified trajectories. In contrast, verified scaffolds remain effective even when the teacher is conditioned on the student's own unsuccessful rollout. Based on this finding, we introduce OASIS, which retains the OPSD objective but supervises mostly verified by label on-policy trajectories and replaces written solutions with unverified model-generated attempts as the teacher context. OASIS therefore requires only final-answer labels. Across Qwen3-1.7B, 4B, and 8B on AIME 2024, AIME 2025, and HMMT 2025, OASIS improves over the base model by 3.2–3.8 points on average, while OPSD's gain falls from 3.05 points at 1.7B to 0.14 at 8B. At 8B, OASIS improves over OPSD by 3.05 points, showing that verified on-policy scaffolds preserve the effectiveness of self-distillation as models scale.

Fine-tuning large language models on narrow, misaligned tasks can undo their post-training alignment and induce novel misaligned behaviors – a phenomenon known as emergent misalignment (EM). EM has been linked to persona-like representations, where fine-tuning might reduce loss by amplifying a harmful or 'evil' persona. It remains unclear which properties of the training data drive this effect: whether all harmful examples contribute approximately equally to misalignment and whether different models are equally affected by the same fine-tuning examples. In this work, we use training data attribution to quantitatively estimate how much each harmful example contributes to EM. We benchmark the quality of the attribution via retraining – a sound attribution score should enable us to enhance or attenuate EM by filtering data on that score. Score-based filtering can substantially enhance or attenuate EM; we find that both data-attribution scores and a black-box harmfulness score can identify consequential examples. All models we test become misaligned when trained on the same dataset, and influence scores perform best when filtering data from the same model that computed them. We find cross-model generalization of influence scores from scores derived from the three model families we tested, but this generalization does not recover same model filtering performance.

Asynchronous Byzantine Reliable Broadcast (BRB) is a fundamental primitive that guarantees agreement and validity in distributed systems subject to Byzantine faults, but it lacks ordering guarantees. Causal message ordering is important for many applications such as blockchain and social networking. Existing solutions for Byzantine Causal Reliable Broadcast (BCRB) have several drawbacks. Such protocols typically append vector clocks or dependency barriers to application messages, resulting in a metadata overhead that scales linearly with $n$, the number of processes in the system. In this paper, we propose the first optimal message overhead BCRB algorithm that guarantees safety. We do this by re-engineering Bracha's BRB algorithm with relatively small but critical modifications, and prove that our algorithm solves BCRB with optimal message metadata. The algorithm achieves constant-size $\mathcal{O}(1)$ message metadata overhead and $\mathcal{O}(n^2)$ messages, resulting in $\mathcal{O}(n^2)$ communication word complexity. This is as against $\mathcal{O}(n^3)$ communication word complexity of existing protocols. The algorithm tolerates $f < n/3$ Byzantine processes, which is the well-known optimal resilience bound, and uses four phases.

We show that low discrepancy is equivalent to spectral gap for directed $k$-uniform hypergraphs. For undirected hypergraphs this recovers a theorem of Lenz and Mubayi, with a considerably shorter proof. At the heart of our argument is a Frieze–Kannan type spectral regularity lemma for hypergraphs based on the variational notion of hypergraph eigenvalues by Friedman–Wigderson, which decomposes any tensor into a bounded number of rank-one tensors plus a quasi-random tensor with small top eigenvalue. This lemma may be of independent interest, and we prove it for complex-valued, not necessarily symmetric tensors. We also briefly discuss a Szemerédi-type variant and the regularization of Cayley-type hypergraphs over finite abelian groups.

Scientific queries are often brief, while relevant papers use specialized vocabulary. Generated query expansion can bridge this mismatch, but earlier work suggests that its value shrinks as the underlying retriever becomes stronger. We test the four generated formats of term lists, a pseudo-document, multiple pseudo-references, and corpus-steered text all together with SPLADE-v3 on NFCorpus, TREC-COVID, and SciDocs. Every condition searches the same frozen document index and follows the same query-side integration rule and 256-dimension budget, isolating the effect of the added content. All twelve method-collection comparisons improve aggregate nDCG@10, with best relative gains of 4.81%, 8.92%, and 9.47%. Eleven remain significant after Holm correction. The gain persists in 103 of 114 interpolation settings, including every setting that assigns at least 30% of the mixture weight to the original query. Shuffled-text and non-contextual lexical-bag controls also remain above baseline in all 24 aggregate comparisons, showing that the added vocabulary carries most of the benefit. A corpus-induced typed concept graph, by contrast, produces no consistent gain, and its relation, depth, validation, random, and gating controls do not rescue it. Generated vocabulary can therefore complement a strong learned sparse retriever, provided that the original query remains strongly represented.

Isolating vocalizations from noise in bioacoustics recordings is a prerequisite to many ecological analyses, including species identification, animal communication understanding, and individual or population-level variability studies. However, when recordings are acquired in open environments, vocalizations, noise, or signal-to-noise ratio can vary widely across individuals, species, environment, and recording conditions, making it hard to develop robust and generalizable methods for ecological analyses. To account for these challenges, noise reduction techniques are used to remove noise before downstream analyses. Popular methods such as Noisereduce rely on spectral gating, which estimates a noise threshold for each frequency channel. We propose a further improvement to Noisereduce, leveraging the fact that most animal sounds have structure across multiple frequencies. We apply our method on bird and marine mammals recordings and show that our extension of Noisereduce leads to improved denoising in both above and under water acoustic recordings, without impacting speed of preprocessing.

Superflares are frequently observed on magnetically active stars, particularly on active M-dwarfs, thereby influencing the habitability of nearby exoplanets. However, the causes of superflares remain unclear. Here, combining FAST and TESS, we detected a finely structured decimetric radio burst appearing as groups of spikes during a white-light superflare (total bolometric energy $\sim 10^{33} \ \mathrm{erg}$) on the M-dwarf binary EQ Peg on September 8th, 2022. The most significant ones appeared near the peak time of the optical flare and lasted for about $100 \ \mathrm{s}$. The radio spikes present extremely high brightness temperatures ($\gtrsim 10^{13} \ \mathrm{K}$) and high degrees of circular polarization ($\sim 0.84 \pm 0.16$), indicating the origin of a coherent emission process, most likely the electron cyclotron maser. Based on Zeeman-Doppler Imaging measurements of EQ Peg, we further locate the sources of the radio spikes, including their heights and latitudes, under the assumption of a large-scale dipolar magnetic field configuration. The results provide strong evidence for finely structured decimetric radio bursts generated by stellar flares and shed light on the origin of superflares.