Computation and Language
Papers
Fresh from arXiv, typeset properly. Highlighted words trade on the market.
All of arXiv
Thursday 27 August
We study audio-visual conflict as a compositional generalization test for AV-LLMs: the model must combine synchronized but semantically incompatible audio and video evidence and decide whether the pair matches. On VideoLLaMA 2-7B-AV, three alignment configurations remain nearchance on the scored exact-string Yes/No subset of AVHBench, even though their output priors shift substantially. Similarly, off-the-shelf InternVideo2 experienced a 32.3% accuracy decrease specifically under cross-modal conflict, accompanied by a 17.3% instruction-following failure. We call this failure mode prior dominance: late-layer commitment to an internally preferred answer pattern that is weakly grounded in the conflicting inputs. To explain this behavior, we conduct a mechanistic interpretability analysis and find that commitment remains concentrated at 25.5 $\pm$ 1 layers. We show that stronger temporal alignment changes answer bias, but do not improve compositional conflict resolution. Code and data to reproduce our mechanistic audit and behavioral evaluations are available at https://github.com/AdarshSudheer09/AVHBench-dmai.
Theoretical Economics
Strong Observable Substitutability and the Cumulative Offer Mechanism
We study the properties of the cumulative offer mechanism (COM) when institutions' choice functions satisfy strong observable substitutability (strong OS). First, we show that each of the following properties of the COM characterizes strong OS: IR monotonicity, weak Maskin monotonicity, and dropping monotonicity. Dropping monotonicity is a new condition weaker than weak Maskin monotonicity, and it is interpretable as a weakening of strategy-proofness and non-bossiness. Second, we show that when choice functions are strongly OS, weak group strategy-proofness of the COM reduces to individual strategy-proofness. However, strong OS is neither necessary nor "almost necessary" for this reduction.
Read the full abstract
Audio and Speech Processing
SURE-Voice: A Front-End Baseline for Speech-Evidence Filtering in Speech LLMs
Speech language models (speech LLMs) can generate plausible outputs from audio that contains no usable speech evidence. We study this failure as a pre-generation support-estimation problem and present SURE-Voice, a training-free front end that decides whether an audio prompt contains intelligible speech evidence before calling a speech LLM. We build SURE-Challenge with a 640-example SURE-Core split and a 1,920-example SURE-Extended split derived from 120 LibriSpeech source utterances. Using one fixed operating point, an energy screen plus Whisper token confidence raises unsupported accuracy on the held-out Extended test from 0.000–0.133 to 0.919 for six non-degenerate speech LLM backbones, while supported accuracy remains 0.919–0.970 and downstream calls fall from 480 to 287. A 500-clip ESC-50 sanity set shows the same pattern on real environmental audio, with vocal non-speech as a residual failure mode. An overlap diagnostic shows that source attribution remains separate from speech-evidence filtering. The evidence supports a controlled benchmark baseline and a deployment-oriented analysis; it does not establish universal robustness to semantic answerability, gain variation or natural conversations.
Read the full abstract
Cryptography and Security
Memorization Is Not Extraction: Tight Differential-Privacy Bounds and Audit Blind Spots
Memorization in large language models is measured through a zoo of definitions whose formal relations are unknown, and differential privacy (DP) is treated as a proxy against all of them at once. We pin down the exact DP constant for the two that carry the practical weight, counterfactual memorization and adaptive extraction, and show that they do not control each other. Under $f$-DP, every adaptive extraction protocol with list budget $m$ succeeds with probability at most $1-f(\kappa)$ for the oblivious baseline $\kappa$, and the bound is tight on a dense set of baselines: DP uniformly controls extraction exactly up to a threshold in how well the secret can be guessed a priori. Min-entropy certifies that baseline distribution-free, since $H_\infty\ge\epsilon\log_2 e+\log_2(m/\tau)$ holds extraction below a risk level $\tau\le1/2$ under pure $\epsilon$-DP for every prior, and is exact on uniform priors. On the memorization side, $f$-DP caps the counterfactual memorization of any bounded score at an advantage functional $\eta(f)$, equal to $\tanh(\epsilon/2)$ under pure DP; for $k\ge2$ duplicated copies the naive $\epsilon\mapsto k\epsilon$ bound $\tanh(k\epsilon/2)$ is unattainable, the exact constant being a closed-form staircase attained by geometric noisy counting. That cap is attained inside the local score class used in practice, and it is there that the two measures separate: one mechanism is memorized yet unextractable, another fully extractable yet exactly invisible to every loss-based score. The two-sided blind spot this opens for loss-based auditing and unlearning verification survives on billion-parameter models: a reserved-trigger release is recovered verbatim from one prompt while the audits practitioners deploy certify it clean.
Read the full abstract
Nuclear Theory
Nuclear lattice effective field theory as a testing ground for $\alpha$-cluster structures in ${}^{24}\mathrm{Mg}$
The framework of nuclear lattice effective field theory (NLEFT) is applied to $^{24}\mathrm{Mg}$, with the perspective of obtaining a model-independent density map of the geometry of a sample of excited states of the nucleus. The Hamiltonian incorporates Wigner SU(4)-symmetric nuclear forces as well as the Coulomb interaction. The coupling constants of the spin-isospin symmetric nucleon-nucleon potentials have been adjusted in order to reproduce the experimental ground-state (g.s.) energy of $^{24}\mathrm{Mg}$ as well as the experimental Tjon ratio between the binding energies of $^3\mathrm{H}$ and $^4\mathrm{He}$. The ensuing parameter set turns out to be capable of capturing the experimental trend of the binding energy per nucleon, reproducing simultaneously within 1% deviation the measured values for $^{18}\mathrm{F}$, $^{22}\mathrm{Na}$, $^{26}\mathrm{Al}$, $^{28}\mathrm{Si}$, $^{30}\mathrm{P}$ and $^{32}\mathrm{S}$. Considerations based on the convergence rate of Euclidean-time extrapolations for the two lowest energy eigenvalues highlight the dual nature of the $0_1^+$ and $2_1^+$ states, of hybrid mean-field and $\alpha$-cluster type. For the latter, triaxial $\alpha$-cluster configurations seem to be favoured over the axially-symmetric ones, whereas oblate superdeformation might characterize a rotational band at $20$ MeV excitation energy.
Read the full abstract
Dynamical Systems
Higher Order Structure along Nilpotent Eigendirections in Planar Vector Fields
We study a distinguished one dimensional restriction associated with planar vector fields whose linearization is a nonzero nilpotent matrix of index two. This restriction admits an intrinsic interpretation on the nilpotent eigendirection, while its scalar representation in Jordan coordinates satisfies a simple transformation law under admissible changes of Jordan chain. Its quadratic Taylor coefficient coincides with the intrinsic quadratic Bogdanov–Takens coefficient. % For rational restrictions, we show that the Taylor coefficient sequence satisfies a finite linear recurrence determined by the minimal denominator. Its degree defines a recurrence degree that is invariant under admissible changes of Jordan chain and gives the minimal order of an eventual homogeneous recurrence. Exact cancellations may therefore be detected directly from the restriction before nonlinear normal form transformations are performed. Applications to four planar models exhibit different recurrence degree and cancellation mechanisms, illustrating the complementarity between the distinguished restriction and smooth nilpotent normal form theory.
Read the full abstract
Astrophysics of Galaxies
High Dispersion Measure Fast Radio Bursts from Young, Dust-Poor H II Regions
Several repeating fast radio burst (FRB) sources exhibit host-frame dispersion measure (DM) excesses far exceeding those expected from the interstellar medium in their host galaxies, implying a substantial contribution from the sub-kiloparsec-scale local environment. These excesses are often attributed to an expanding supernova remnant; however, this canonical interpretation does not explain their apparent preference for low-metallicity dwarf galaxy hosts, suggesting that the mechanism producing the local DM is coupled to the properties of the host galaxy. In this work, we investigate whether the H II regions generated by young, massive stellar associations can simultaneously explain these DM excesses and their preferential association with low-metallicity dwarf galaxies. We develop a semi-analytic model for FRB-emitting neutron stars in cluster-forming molecular clouds and perform idealized one-dimensional radiation-hydrodynamic simulations using time-dependent ionizing luminosities from pySTARBURST99. Our findings indicate that high-mass $(M_\star \sim 10^5-10^6\;M_\odot)$, young ($\lesssim 10\;{\rm Myr}$) stellar associations in low-metallicity environments (LMC-like or lower) can produce ${\rm DM}_{\rm HII} \gtrsim 10^3\;{\rm pc\;cm^{-3}}$ for FRBs that form early ($\lesssim 5$ Myr) and remain embedded in the ionized gas. The low-mass/metallicity preference arises from dust: absorption of the Lyman continuum suppresses the ionized column at higher metallicity, making extreme local DMs preferentially generated in dust-poor, low-metallicity environments. At cosmological distances, selection of a metal-rich, massive, and quiescent host galaxy sample where such extreme local contributions are likely to be suppressed may yield a cleaner FRB sample for using extragalactic DMs as tracers of the intervening intergalactic and circumgalactic media.
Read the full abstract
Computational Engineering
Investigating Forecast Proficiency of Hurricane-Induced Compound Flooding With a Discontinuous Galerkin Shallow Water Equation Solver
Recent severe storms on the U.S. Gulf coast have demonstrated the challenges presented by compound flooding, such as the interactions between rainfall runoff and storm surge. Historically, many studies have neglected these nonlinear interactions, but we propose to use a discontinuous Galerkin shallow water equation solver, which allows for incorporation of rainfall inputs directly onto the finite element mesh. In this work, we analyze the use of parametric rainfall for forecasting scenarios, using Hurricane Beryl (2024) as a case study. Beryl led to extensive flooding due to rainfall and storm surge along the Gulf. We use a collection of the National Oceanic and Atmospheric Administration's short-term advisories along with the best track data to demonstrate the efficacy of the parametric rainfall model for forecasting. Results show that the parametric rainfall input allowed for much more accurate inundation. Areas with heavy rainfall and low surge were affected the most, with many areas peaking over 50 cm above the baseline surge model. Almost none of the available high water marks from Beryl were captured by the standard models, but the compound flooding models capture many of them, the majority of which show relative errors under 10 percent. Results from the advisory forecast simulations were shown to be much closer to the best track hindcast simulation when rainfall forcing was used, even while early forecasts predicted the storm's trajectory much less accurately. The advisory simulations improved even further as Beryl neared the Texas coast, with sampled peak elevations most closely approximating the best track at Advisory 38, a few hours before landfall.
Read the full abstract
Computational Geometry
Optimal exponential memory for sequential Euclidean connections: edge-power costs and phase transitions
We study the edge-power cost of the labelled tree generated by the $\gamma$-strategy, a constant-gain rule for sequential Euclidean connections. Starting with $x_0=p_0$, each input point $p_i$ is attached to $x_{i-1}$, and the state is updated by $x_i=\gamma x_{i-1}+(1-\gamma)p_i$. Retaining $x_i$ subdivides the insertion segment into a spine edge and a leaf edge. The memory parameter $\gamma$ controls how long earlier points influence subsequent attachment points. We minimize the sum of the $\alpha$-powers of the edge lengths under independent uniform input and arbitrary input sequences. For uniform points in the unit ball, the stationary problem has a transition at $\alpha=1$. Its continuous extension is minimized at the boundary for $0<\alpha\leq1$, while every global minimizer is interior for $\alpha>1$. We determine the finite optimizer in the joint window $\alpha_N=1+\varepsilon_N$, $\varepsilon_N\log N\to\lambda$. Below an explicit threshold it lies on the $N^{-1/2}$ scale, at the threshold its scale is $\sqrt{\log N/(N\log\log N)}$, and above the threshold it approaches an explicit stationary root with two computable corrections. A second threshold identifies the governing correction, and differentiated estimates prove eventual uniqueness. At $\alpha=3d+8$, the linear coefficient at the stationary endpoint changes sign and a branch of strict local maxima enters the parameter interval. For arbitrary input sequences, the optimal parameter and asymptotic worst-case edge-power cost per point are explicit for $0<\alpha\leq3$. At high powers, periodic antipodal block inputs give explicit lower bounds which, with a separation argument, show that the optimized cost is asymptotic to $2\log2/\log\alpha$. Exact results for powers two and four, a rational recursion for every even power, and a high-dimensional expansion complete the analysis.
Read the full abstract
Optimization and Control
Optimal control of fractional diffusion with Dirac measures
We study a PDE-constrained optimization problem for an elliptic equation with the spectral fractional Laplacian and a linear combination of Dirac measures as the forcing term; the controls are the amplitudes of these singular sources. We prove existence and uniqueness of an optimal solution and derive first-order optimality conditions. We then propose a discretization based on finite elements. Since the set of admissible controls is finite dimensional, the control variable itself does not require discretization. We conclude by deriving a priori error bounds
Read the full abstract
Instrumentation and Methods for Astrophysics
PolPy: A universal tool for X-ray/gamma-ray polarimetry using common data format standards
X-ray/gamma-ray polarimetry has been a rapidly developing field in recent years. However, the lack of standardized analysis frameworks has hindered cross-instrument comparisons and joint multi-mission studies. To address this, we present PolPy, a universal tool designed for polarimetry of high-energy astrophysical transients. Built as a plugin for the Multi-Mission Maximum Likelihood (3ML) framework, PolPy introduces standardized data and response formats, common reference frames based on IAU conventions for defining polarization angles, and a robust likelihood-based template-matching algorithm. The article briefly describes these standards and presents tests and results performed to validate the software using simulated gamma-ray burst (GRB) injections across two different mission mass models: POLAR and Daksha. Our verification tests demonstrate that PolPy accurately recovers the injected polarization angles and fractions without systematic bias. Furthermore, we show that performing simultaneous joint fits across multiple detectors and instruments can significantly improve parameter constraints compared to single-instrument analyses. By unifying data formats and methodology, PolPy enables reliable joint polarimetric analysis for current and future high-energy polarimeters.
Read the full abstract
Optimization and Control
Beyond Procrustes distances: a multilinear Gromov-Wasserstein distance capturing chirality
Efficiently and robustly analyzing shape data is critical across many scientific disciplines. While chirality is a fundamental property in numerous applications - most notably in molecular science - existing shape analysis metrics either fail to distinguish enantiomers, or rely on costly alignment. To address this gap, we introduce a multilinear generalization of the Gromov-Wasserstein objective, which yields an alignment-free distance between shapes, represented as probability distributions quotiented by a Lie group $G$. In particular, for $G = SO(d)$, we introduce the Chiral Gromov-Wasserstein ($\mathrm{CGW}$) distance, sensitive to chirality. We establish robustness and sample complexity properties for the multilinear Gromov-Wasserstein distances and develop efficient algorithms to compute them, reformulating the underlying optimization problem by projecting couplings onto a low-dimensional space. We derive algorithms for both local and approximate global solutions, yielding an approximation scheme, which is polynomial in the sample size and precision. We validate the framework through numerical experiments that demonstrate the effectiveness of $\mathrm{CGW}$ as a shape metric for chiral objects.
Read the full abstract
Astrophysics of Galaxies
Near-Infrared Spectroscopy of Low-Redshift Palomar-Green Quasars and Composite Spectra
The Palomar-Green (PG) Quasars have been used extensively to establish the spectral properties of luminous active galactic nuclei (AGNs) at low redshifts in many wavebands, but not the near-infrared (NIR). We present NIR spectra of all 87 PG quasars with z < 0.5 observed with the Apache Point Observatory 3.5-m telescope and the TripleSpec spectrograph in standard cross-dispersed mode. Here we employ our data set to construct a NIR composite quasar spectrum spanning rest-frame wavelengths from approximately 8800-22000 angstroms reaching a signal-to-noise ratio of several hundred in the H-band for 4 angstrom pixels. We characterize the continuum and emission-line properties of this composite. Additionally, we construct composite spectra as a function of luminosity and Eddington ratio, finding some differences, not entirely unexpected. In particular, more luminous composite NIR spectra grow increasingly red at long wavelengths, consistent with relatively stronger dust emission at higher temperatures.
Read the full abstract
Systems and Control
Design of a Fast Crowbar Protection Circuit for Single Event Effect Testing
We describe the design of a custom crowbar circuit that quickly removes power when a programmable current threshold is exceeded. This protects devices from permanent damage during single event latchup/burnout tests.
Read the full abstract
Geometric Topology
A normal form for the Thin Flat Surfaces category
The category TFS of thin flat surfaces, introduced by Khovanov, Qi, and Rozansky, is a strict symmetric monoidal category whose morphisms are compact oriented surfaces with corners arising as neighbourhoods of immersed graphs in a strip. We prove two main results. First, a normal form theorem: every connected viewable tf-cobordism is equal, in TFS, to a canonical composite determined by its topological type. Second, a sufficiency theorem: we establish a list of relations in the TFS category which are sufficient, meaning that any two composites of generators representing the same tf-cobordism are related by a finite sequence of the listed relations. Together, these results give a complete generators-and-relations presentation of TFS.
Read the full abstract
Statistical Methodology
Accelerating Bayesian Variable Selection using Piecewise Deterministic Markov Processes
Bayesian variable selection becomes computationally challenging when models contain many dependent parameters. We study Piecewise Deterministic Markov Process (PDMP) samplers as a continuous-time alternative to conventional Markov chain Monte Carlo for spike-and-slab variable selection. In sticky PDMP samplers, active parameters evolve continuously until they reach zero, where they may remain for a random duration before re-entering the model. While one parameter enters or leaves the model, the remaining parameters continue to evolve along the deterministic flow, offering a potentially advantageous mechanism for exploring posteriors with strongly dependent parameters. We make two methodological contributions. First, we extend existing sticky PDMP methods beyond independent spike-and-slab priors to dependent model priors and dependent slab distributions. Second, we investigate the use of unbiased stochastic gradients to reduce the computational cost of variable selection when the likelihood decomposes into many factors while retaining the same target distribution. We study these extensions to two psychometric models: a Gaussian random intercept cross-lagged panel model and an ordinal Markov random field. For the former, marginalization yields a fixed-dimensional sufficient-statistic representation that permits efficient model evaluation. For the latter, the model factorizes, which enables subsampling over person-node contributions. In simulation studies, we compare ZigZag, Bouncy Particle, and Boomerang dynamics with reversible-jump MCMC, and examine the effects of prior dependence and stochastic-gradient subsampling on sampling efficiency. We illustrate the methodology using data from an empirical study on mental well-being. Finally, we discuss the advantages and challenges when using PDMP samplers for Bayesian variable selection.
Read the full abstract
Algebraic Geometry
The Segre invariant and Brill-Noether theory on Hirzebruch surfaces
We study the Segre invariant on rank two $H$-stable vector bundles with fixed Chern classes on Hirzebruch surfaces $\Sigma_{n}$. For each vector bundle on these surfaces, we determine the possible values of the Segre invariant and compute the number of global sections of these bundles and present applications to Brill-Noether Theory.
Read the full abstract
Artificial Intelligence
Why Didn't It Check? Unsupported Final Claims and Their Repair in Two Tool-Equipped Language Models
A language model with access to tools can commit to a final claim unsupported by the evidence it has seen, even when a single available tool call would resolve the uncertainty and its instructions explicitly forbid assumptions and guesses. We separate this failure into two precisely defined quantities: occurrence, how often the model makes an unsupported claim on its own, measured from the visible evidence and final claim without using the hidden correct answer; and conditional repair, how often those same naturally occurring unsupported claims are repaired when the missing evidence is supplied. On one fixed Qwen3-32B setup, 33 of 512 first responses to 256 new prompt templates ended with an unsupported established claim. We replayed each case from an exact copy of the state in which the claim occurred; within each matched replay, the alternative tool responses had the same structure and length and differed only in a one-character response code. Resolving evidence repaired 33 of 33 claims; a matched response carrying no useful information repaired 0 of 33. When the evidence supported the original answer, the model preserved 33 of 33, with no observed harm. In a separate experiment, on 64 cases where evidence was needed, an automatic checking rule added 21 evidence calls, corrected all 10 wrong unsupported claims, preserved the 11 that were correct by accident, and never changed a correct answer into a wrong one. On a fixed Gemma 4 setup using the same sampling settings, the model called the tool in all 512 first responses and never made an unsupported final claim, so conditional repair could not be measured for that setup. These results describe two local fixed model setups on two synthetic task families. They do not show how common this failure is in real-world deployments, nor that it reflects a general mechanism shared across models.
Read the full abstract
Mathematical Physics
Fractional Hall Conductance of Laughlin States from a Topological Index
We provide a rigorous Laughlin-pump argument that establishes the fractional quantization of the Hall conductance directly from the Laughlin state as a topological index. Our proof uses a recently defined index of a pair of pure states together with an infinite matrix product state analysis that is rigorous on a sufficiently thin cylinder. By virtue of the connection to an index, our result implies, in particular, that Laughlin states are representatives of topologically stable fractional quantum Hall phases.
Read the full abstract
Cryptography and Security
Revisiting Continuous Noise Sampling for Multi-Party Differential Privacy
Combining secure multi-party computation (MPC) with differential privacy (DP) enables multiple parties to release aggregate statistics without a trusted curator, and the core primitive is the protocol to sample noise from a continuous distribution under finite-precision arithmetic. In this paper, we revisit the continuous noise sampling protocols and present several improvements in both security and efficiency. We start by identifying a vulnerability in widely used sample-and-scale constructions. We demonstrate that the scaling operation in arithmetic circuits confines the noise to a sparse, publicly known set of values, so that an adversary can observe the released noisy queries and decide which dataset produced them. As concrete demonstrations, we instantiate attacks on two systems employing such “flawed” sampling protocols: Orchard (OSDI'20) for DP secure aggregation and DP-BREM$^+$ (USENIX Sec'25) for DP federated learning. We report a near-$100\%$ attack success rate on both systems, under any noise scaler $s\geq 2$ used in practice. The leakage we reveal is intrinsic to the scaling operation, and direct repairs either substantially sacrifice utility or add significant precision bits to make the sampling more expensive. To address the security and efficiency issues together, we turn to discrete sampling at the granularity of individual biased bits. We make several optimizations to the sampler and prove its security. Our implementation achieves $4\times \sim 612\times$ speedup over existing secure discrete samplers and orders-of-magnitude speedup over the insecure sample-and-scale paradigm, with negligible utility loss compared to the ideal continuous mechanism.
Read the full abstract
Instrumentation and Detectors
Frequency Synchronization Circuit Model for Analog Vector-Matrix Multiplication in the Frequency Domain
This work introduces a frequency synchronization circuit model for analog vector-matrix multiplication in the frequency domain. The frequency synchronization is implemented by coupling oscillators through all-to-all connections. Each oscillator is implemented using complementary cross-coupled transconductance pairs and RC high-pass filters. In this work, the frequency synchronization is achieved not only using oscillators alone but also through a configuration in which oscillators and resonator models (i.e., MEMS/NEMS resonators) are combined. While addressing the frequency synchronization of all the coupled oscillators, a partial synchronization is also discussed by controlling the coupling strength of a selected subset of oscillators. This frequency synchronization analysis focuses on the small-signal analysis using analog circuit and resonator models, thereby enabling an understanding of the synchronized frequency defined by the parameters of the analog circuit and resonator models.
Read the full abstract
Quantum Physics
Benchmarking Quantum Feature Encoding Strategies for Binary Classification with QSVM
The way in which classical data are encoded into quantum states plays a significant role in both classification performance and quantum circuit complexity in Quantum Machine Learning. In this study, the effects of different quantum feature encoding strategies on Quantum Support Vector Machine performance were investigated using five binary classification datasets. In particular, the statistical relationships between features were incorporated into quantum circuits through \(RY(\theta)\) and controlled-\(RY(\theta)\) gates, and this approach was compared with conventional quantum feature maps. The results demonstrate that incorporating statistical relationships into the encoding process can influence classification performance. However, more complex and densely entangled circuits do not necessarily yield higher performance. In addition, a composite evaluation metric was employed to jointly assess predictive performance, generalization, and circuit cost. The findings across the five datasets indicate that the choice of quantum feature encoding strategy should account for the underlying structure of the data and that predictive performance should be evaluated together with quantum circuit complexity.
Read the full abstract
Machine Learning
Fast Weight Attention for Continual Learning
Recurrent fast-weight memories and selective state-space models compress an expanding context into a fixed-size recurrent state, making the state transition an online learning rule. We study this rule under read-after-write autoregressive semantics. For the prefix-prediction objective considered here, the local fast-memory example revealed at step $t$ is the prefix-aligned pair $(\mathbf{x}_t,\mathbf{y}_t)=(\phi(\mathbf{k}_{t-1}),\mathbf{v}_t)$. The common same-step association $(\phi(\mathbf{k}_t),\mathbf{v}_t)$ remains causal, but optimizes a different internal objective. We derive normalized first-order updates for squared-error regression and negative inner-product objectives. The regression family comprises Falcon-1 (a scalar NLMS update), Falcon-2 (its per-column extension), and Falcon-3 (a sliding-window mini-batch update); Falcon-1A/Falcon-2A/Falcon-3A are the corresponding inner-product variants. We provide recurrent, masked-parallel, and chunk-parallel forms, together with numerically stable positive-decay renormalization. Representative variants remain competitive in language modeling and improve length extrapolation on variable-digit addition. This framework separates temporal alignment, plasticity, forgetting, and bounded rehearsal in recurrent sequence models.
Read the full abstract
Rings and Algebras
Integral Orders for the Okubo Algebra and Idempotent Geometries of the $E_8$ Lattice
We study the Coxeter-Dickson $E_8$ order as a common integral support for the octonionic, para-octonionic and compact real Okubo products. The three algebra structures have the same additive lattice, the same positive composition norm and hence the same $240$ norm-one vectors, namely the roots of $E_8$ and the vertices of the Gosset polytope $4_{21}$. Their multiplicative structures are nevertheless different and this difference is already visible in the idempotents: among the same $240$ roots one finds respectively $1$, $57$ and $12$ nonzero integral idempotents. We show that these three counts organize the common root system in three increasingly polarized ways. The octonionic product leaves the full $E_8$ geometry unseparated; the $57$ para-octonionic idempotents reproduce the $E_7$ contact decomposition $1+56+126+56+1$; the $12$ Okubo idempotents split into four mutually orthogonal oriented $A_2$ triangles and hence determine canonically an $A_2^4$ subsystem. Choosing one triangle as the external $A_2$ then yields the $A_2+E_6$ Magic Star, while the remaining nine idempotents determine the trinification subsystem $A_2^3\subset E_6$. This provides an arithmetic interpretation of the $E_8\supset E_7$ and $E_8\supset E_6+A_2$ decompositions directly from integral idempotents.
Read the full abstract
Algebraic Geometry
$K_2$-regularity is equivalent to smoothness for toric varieties
Recent work by Wanchun Shen shows that $K_2$-regular affine local complete intersections of any dimension in characteristic zero are smooth. In this note, we show that the same holds for affine toric varieties.
Computation and Language
Informational Antilocality and the Locality Bias in LLMs
We consider the ability of transformer-based language models (LLMs) to learn what we call k-antilocal languages, i.e., languages that have no mutual information across any span of $k$ contiguous symbols. We construct such languages with increasing $k$, finding that LLMs trained on them achieve comparable cross-entropy loss regardless of antilocality, but converge more slowly on more antilocal languages. Our findings support the idea that non-local dependencies are more difficult to learn, but the evidence for this bias comes from learning speed rather than learning success.
Read the full abstract
Signal Processing
Fast Tri-Hybrid Beamforming via Deep Unfolding
Tri-hybrid multiple-input multiple-output architectures have recently emerged as a promising enabler for next-generation wireless systems, as they potentially provide enhanced design flexibility without a proportional increase in hardware cost or power consumption. However, the resulting triple-domain coupling renders beamforming optimization challenging and computationally demanding. This paper develops a fast tri-hybrid beamforming framework for multiuser downlink systems employing dynamic metasurface antennas (DMAs). Based on the equivalence between weighted sum-rate maximization and weighted sum-minimum mean square error minimization, an iterative algorithm is first derived under per-DMA input power constraints, with all update equations available in closed form, but convergence inherently remains slow. To enable real-time operation, the algorithm is further unfolded into a trainable finite-iteration architecture using graph neural networks that ensure permutation equivariance and support varying numbers of users. Trained on ray-tracing channel data, the unfolded method achieves comparable or higher system sum-rates while reducing runtime by more than an order of magnitude. The method also demonstrates strong scalability, robustness, and generalization across various environments.
Read the full abstract
Databases
Real-time SQL Plan Management in Oracle
Consistent query performance is essential for mission critical database applications, yet SQL execution plans can change due to factors such as database upgrades, DML changes, new indexes, etc. While plan stability mechanisms such as stored outlines prevent regressions by freezing execution plans, they also inhibit performance improvements by disallowing plan evolution. We introduced SQL Plan Management (SPM) in Oracle 11g to address this trade-off by maintaining a set of accepted execution plans and allowing plan evolution only when new plans demonstrably outperform existing baselines. However, prior implementations of SPM primarily rely on background performance verification processes, delaying regression detection and recovery. This issue is amplified in autonomous cloud database systems, where several automatic actions that could cause plan change driven regressions are performed with limited customer control. Timely detection and remediation is paramount, but the constrained background resources on cloud may not keep pace. To overcome these limitations, we introduce Real-Time SPM in Oracle 26ai, a novel extension of SPM that performs foreground verification of new execution plans during user query execution. Real-Time SPM leverages runtime session context to immediately validate plan changes, enabling rapid adoption of superior plans while promptly detecting and preventing regressions. This paper presents the architecture and design of Real-Time SPM - including technical challenges like reliably comparing performance of previous plans - and contrasts it with traditional background plan evolution.
Read the full abstract
Machine Learning
Beyond Search-Imitation: Prior-Directed Exploration for Searchless Chess
Searchless chess networks reach human master strength from a single forward pass by imitating a stronger teacher: the strongest, Leela Chess Zero's (Lc0) released Chessformer, distills the visit counts of an AlphaZero-style Monte Carlo Tree Search (MCTS). Imitating a search is a poor proxy for playing without one, so we fine-tune for single-pass strength with self-play reinforcement learning (RL). Its exploration is usually supplied by an entropy bonus, the reverse Kullback-Leibler (KL) divergence to uniform. We replace it with a forward, mass-covering KL toward the network's own MCTS prior (prior-directed exploration), so exploration covers the moves the prior judges promising, and pair it with an entropy-adaptive sampling temperature, set by the value head's outcome uncertainty, that sharpens once a position is decided. In about two thousand steps it raises puzzle accuracy from 93.9% to 94.9% on a 100,000-puzzle suite and mate-in-four accuracy from 77% to 81% while holding searchless strength at or slightly above the base. Measuring tactical accuracy and playing strength together across a matched-compute sweep, we find the two dissociate: accuracy gains fall in a one-point band while ratings straddle the base, and a control fine-tuned on puzzles alone posts the study's largest tactical gains while shedding roughly 260 Elo; a better puzzle-solver is not thereby a stronger player. Distribution-level measurements show what anchoring buys: without a regularizer self-play collapses onto a single line of play, and the puzzles newly solved are the near misses whose winning move the prior kept alive. The forward-KL prior tops the rating ladder, statistically tied with a reverse-KL anchor that concentrates twice as hard and drops the hardest solutions the mass-covering prior keeps in support.
Read the full abstract
Computation and Language
Load-Bearing Context: The Question Damage Score for Evaluating Context Reliance in Linguistic Reasoning
Determining whether large language models derive answers from context or prior knowledge remains a fundamental challenge. Self-contained linguistic olympiad puzzles provide a controlled setting where all answers derive solely from expert-designed context examples without external knowledge. Removing individual context examples can eliminate information needed for specific questions while leaving the rest of the puzzle unchanged. We leverage this to introduce a diagnostic framework for analyzing individual context examples. Using 53 UK Linguistics Olympiad puzzles, we generate two modified variants by deleting a single context example: (1) uniform random deletion, and (2) targeted deletion (inspired by error-correcting codes) to remove a structurally load-bearing example uniquely carrying necessary information. We formalize this impact using a Question Damage Score to classify puzzles as fragile or robust. Evaluating three frontier LLMs under instructions to abstain when information is insufficient, we find they rarely abstain, often continuing to produce correct answers after load-bearing context is removed. These findings motivate further investigation into context-based reasoning, prior knowledge, memorization, and linguistic inference. Beyond abstention, the framework enables fine-grained analyses of context reliance, including causal interventions, stopping-set analysis, targeted contamination studies, and mechanistic interpretability.