Computation and Language
Papers
Fresh from arXiv, typeset properly. Highlighted words trade on the market.
All of arXiv
Thursday 27 August
Recent prominent post-training methods, such as Reinforcement Learning (RL) and On-Policy Self-Distillation (OPSD), have driven rapid progress in mathematical reasoning for large language models, yet their reliance on ground-truth labels precludes test-time training (TTT). Replacing ground truth with majority-vote pseudo-labels is a natural alternative, yet it is fragile: an incorrect vote corrupts the teacher and misleads every token. We observe that this failure mode is asymmetric: rollouts that disagree with the pseudo-label are typically wrong regardless of whether the vote itself is correct. Building on this observation, we propose Test-Time Policy Optimization (TTPO), an asymmetric objective that distills agreeing rollouts via OPSD and penalizes disagreeing rollouts with Grouped RL. Token-level selection further refines both branches: distillation down-weights already-converged positions, while RL penalizes only confident errors. Both updates remain well-grounded even under frequent pseudo-label errors, and majority-vote routing yields tighter self-supervision as the model improves. Without any labels, TTPO matches label-supervised OPSD on five competition-level benchmarks, raises Qwen3-1.7B from 38.0% to 45.2% in TTT, yields +25.2% to +36.4% without thinking, and shows strong cross-task generalization.
High Energy Physics (Theory)
Proof of the AGT Conjecture at Generic $\beta$
We give an all-level proof of the four-point $SU(2)$ AGT correspondence with four fundamental hypermultiplets at generic $\beta=-\epsilon_1/\epsilon_2$, by proving an all-level factorization formula for Selberg averages of generalized Jack polynomials. Taking a coefficientwise Jack limit of the generalized Macdonald Pieri rule, we obtain the required one-box matrix elements in the strict Cauchy dual basis. A rational corner-function identity then evaluates the sum over all parent double partitions, while an explicit total derivative of the Selberg kernel yields a triangular recursion on the Dotsenko-Fateev charge balance hyperplane. The unique solution of this recursion is the generalized Kadell formula previously verified only through finite level. Combining this result with the generalized Cauchy identity identifies each double partition term with the corresponding Nekrasov fixed-point contribution.
Read the full abstract
Strongly Correlated Electrons
Exact Stiffness and Dynamical Responses from Fock-Space Fragmentation
Exactly solvable quantum many-body models are rare, and even when their spectra are algebraically organized, dynamical responses generally remain difficult to obtain because they probe an extensive number of excited states. Here we show that quantum geometric nesting (QGN) models admit an unusually strong form of solvability rooted in Fock-space fragmentation: excitations on top of the exact frustration-free ground states decouple into Krylov subspaces with a fixed number of particle and hole operators, and hence remain dynamically invariant. Exploiting this structure, we prove that the stiffness of the spontaneously broken continuous symmetry in QGN models is exactly equal to its variational value in the Gaussian manifold, confirming a conjecture from quantum many-body bootstrap~\cite{GaoHanKhalaf2026}. The proof shows that an infinitesimal phase twist couples the ground state only to the one-particle, one-hole fragment, which coincides with the tangent space of the ground state within the variational manifold, thereby making the variational curvature exact. More generally, perturbations whose action remains within a fixed Fock-space fragment have response functions determined exactly by the corresponding few-body sector, enabling exact access to quantities including static susceptibility, optical conductivity, dynamical structure factors, and single-particle Green's functions.
Read the full abstract
Mesoscale and Nanoscale Physics
Band's Geometry Origin of Quantum Spin Transport Phenomena
We develop a geometric description of spin-dependent transport based on the local geometric structure of electronic bands and the Fermi surfaces. For quasi-two-dimensional systems, we show that hyperbolic regions of constant-energy surfaces generate a geometrical contribution to the Fermi velocity that couples naturally to electron spin and produces a spin-current response. We further show that, in the presence of time-reversal symmetry, the algebra of spin operators can be related to the exterior algebra of the band's tangent space, providing an additional geometric interpretation of spin in momentum space. This framework motivates a symplectic description of spin-separated transport on Fermi surfaces and its extension to three-dimensional band manifolds through contact geometry. Our results establish a direct connection between Fermi-surface geometry and intrinsic spin transport.
Read the full abstract
Mesoscale and Nanoscale Physics
Trapping $e/4$ quasiparticles in bilayer graphene
Measuring the charge of the quasiparticles hosted by even-denominator fractional quantum Hall (FQH) states is essential to identify the topology of their ground state. Here, we use a gate-defined antidot in bilayer graphene, with an additional gate to control only the antidot potential, to measure the charge of the quasiparticles trapped around it in even-denominator FQH states. We observe a localized charge of $e/4$ at $\nu=-5/2$, $-1/2$, and $3/2$, consistent with the minimal excitation expected for leading candidate even-denominator ground states, and $e/3$ at the hole-conjugate state $\nu=2/3$. We further show that increasing the coupling between the antidot-bound states and extended edge states drives a crossover between two regimes, characterized by the minimal-excitation gate-voltage period and approximately twice that period, respectively. We discuss two possible explanations for this crossover: quasiparticle bunching and a crossover between distinct antidot transport regimes. Our results, together with previous observations of the daughter states, show that the even-denominator FQH states in bilayer graphene are compatible with a non-Abelian ground state, and that their quasiparticles can be localized around a quantum Hall antidot, a necessary ingredient for topological quantum computation.
Read the full abstract
Human-Computer Interaction
Do User-Authored Permission Policies Improve Protection Against AI Agent Overreach?
AI agents are poised to become a primary interface to digital products, acting across email, files, payments, and personal data. People without professional software backgrounds need understandable, reusable ways to control actions across services. We examine a mechanism in which a language model maps actions to plain-language consequence categories with user-authored "allow", "ask", or "never" rules. We ask what is gained and lost when decisions are made in advance as reusable rules rather than separately for each action. We analyzed 113 participants without professional software backgrounds across three conditions: per-action human-in-the-loop approval (HITL), automated per-action model review (AUTO), or user-authored consequence policy (POLICY). Participants judged 2 examples in each of 4 consequence categories; POLICY participants then set one rule per category. All supervised an 18-action simulated day, including 7 overreach actions. POLICY blocked less overreach than HITL (-20.1 percentage points, 95% CI [-32.1, -8.1]) and AUTO (-14.5 points, 95% CI [-25.8, -3.2]). POLICY lowered runtime prompts from 18.0 to 10.9, but total intervention time was not reliably lower when rule setup was included. Exploratory analysis showed that participants chose "ask" for 114 of 140 POLICY rules, returning most overreach actions to runtime. Of the 148 overreach actions executed in POLICY, 133 followed human approval and 15 ran automatically under "allow" rules. Across all 7 overreach actions, POLICY had the highest approval rate. Counterintuitively, user-authored rules did not by themselves provide stronger protection: many actions outside users' original requests went through after users approved them. These results reveal a gap between preference and commitment: repeatedly choosing "ask" preserves case-by-case choice but prevents a standing policy from settling decisions in advance.
Read the full abstract
Software Engineering
From Static to Dynamic: Benchmarking Real-World Code Review with MCR-Bench
In real-world software development, code review typically involves iterative interactions between developers and reviewers to improve software quality, making the process costly and time-consuming. Although recent work explores large language models (LLMs) for automated code review, most approaches oversimplify code review into a single-round, static decision task, which fails to capture the multi-round interactive nature and the complex problem-solving processes inherent in realistic review scenarios. To bridge this gap, we introduce MCR-Bench, the first defect state-aware benchmark designed for realistic multi-round code review. MCR-Bench covers five commonly-used programming languages and consists of 2,269 real-world multi-round code review tasks, each of which is annotated with fine-grained defect information and cross-round state labels. Each task in MCR-Bench is equipped with fine-grained defect metadata (e.g., description, type, severity) alongside dynamic state annotations, capturing the complete evolutionary trajectory of a defect throughout the multi-round process. We obtain several findings through extensive experiments on MCR-Bench with mainstream LLMs. (1) Limited overall capability: experiments reveal that mainstream LLMs exhibit limited overall performance in defect detection and defect lifecycle state tracking, with performance degrading significantly as the number of interaction rounds increases; (2) Defect-sensitive performance: LLMs' performance varies substantially across different defect types and severity levels, with semantically complex or low-salience defects being significantly more likely to be missed; (3) Underlying Failure Mechanisms: our in-depth error analysis dissects the distinct drivers of false positives and false negatives, revealing critical weaknesses such as cross-round temporal misalignment and inadequate long-range memory.
Read the full abstract
Algebraic Geometry
Towards the Global Torelli Theorem
In this paper, we prove finiteness results for K-trivial varieties in all dimensions. In dimension 2, these results are corollaries of the Global Torelli Theorem for polarized K3 surfaces. In particular, we establish a sufficient condition when two fibers of a projective family of K-trivial varieties are isomorphic via the Hodge theory. As applications, we prove that if a projective family of K-trivial varieties has a dense subset of birationally equivalent fibers, then all fibers are isomorphic.
Read the full abstract
Quantum Physics
Dynamics of local quantum information in random unitary circuits
The physics of information scrambling in quantum many-body systems is intimately related to thermalisation and emergence of chaos. However, its standard characterisation through bipartite entanglement or operator growth remains inherently coarse-grained, obscuring the spatiotemporal anatomy of how quantum information flows between different regions of the system and across various length scales. In this work, we develop a theory for the dynamics of local information in local random unitary circuits which resolves the fine-grained microscopic structures of local information flow in such systems. Using the framework of information lattice which systematises the information content at each length scale within each subsystem, we identify a distribution of length scales at which information resides and obtain the dynamics of the distribution. For Haar-random circuits with infinite local Hilbert-space dimensions, we map this dynamics onto an exact classical stochastic process which reveals that this distribution is described by a Tracy-Widom form which moves ballistically in time ($ \propto t$) toward larger scales, accompanied by a $\sim t^{1/3}$ broadening. We also find the same qualitative behaviour for random Clifford circuits acting on qubits. The similar scaling behaviour in two rather different settings hints strongly towards the universality of our results. In the case of Clifford circuits, we develop a phenomenological Markov process for the dynamics of the stabiliser generators in a specific gauge which confirms the Tracy-Widom distribution and the $t^{1/3}$ scaling of the fluctuations. Ultimately, our results establish the universal properties of the dynamics of local quantum information in a length scale-resolved fashion and provide a possible route towards bridging exact microscopic theories for information dynamics with emergent hydrodynamic descriptions of information flow.
Read the full abstract
Cryptography and Security
RedEvoAgent: Automatic Red-Teaming Agent with Experience-Driven Skill Evolution
LLM-based agents are increasingly deployed in product-level execution harnesses, where jailbreaks can trigger harmful tool use and persistent state changes, creating greater risks than unsafe text generation alone. Existing automatic red-teaming methods often rely on fixed attacks, while recent agentic attackers coordinate multiple jailbreak tools and show stronger potential through trajectory-based retrieval. However, such retrieval can reuse misleading experiences due to retrieval bias and unclear tool credit, and full trajectories add context overhead while reducing interpretability. We propose RedEvoAgent, a black-box red-teaming agent that distills cross-case attack trajectories into a concise, human-readable attack skill. The attack skill adaptively evolves through tool-effectiveness profiling and Deciding-Tool Attribution for skill updates, and a validation ratchet that retains only updates improving validation performance. Experiments on multiple benchmarks, target models, and target execution harnesses show that RedEvoAgent outperforms fixed and agentic baselines, improves tool efficiency, and transfers across attacker models and target execution harnesses.
Read the full abstract
Combinatorics
Graded Ehrhart theory for hypersimplices
We prove that the $q$-Ehrhart series of a hyperplane slice of a cube is a rational function with an explicit denominator that satisfies $q$-reciprocity, confirming a conjecture of Reiner and Rhoades for these polytopes. To do this, we find a generating set for the orbit harmonics ideal, which also yields the Hilbert series and graded Frobenius characteristic of the associated quotient. We further show that the harmonic algebra of a hypersimplex $\Delta$ is generated as an algebra by the harmonic space of $\Delta$ using structural results on 2-factors of regular multigraphs. In particular, the harmonic algebra is finitely generated, giving a second proof of rationality.
Read the full abstract
Solar and Stellar Astrophysics
The JWST/NIRCam Scattered Light Disks GTO 2780 program: panchromatic coronagraphic imaging of the HD 10647, HD 32297, HD 61005, HD 107146, and HD 181327 debris disk systems
Debris disks, composed of rocks, boulders, planetesimals, and the dust produced in their collisions, present the most readily observable components of mature planetary systems. They also serve as valuable diagnostic tools, enabling studies of planetary dynamical interactions and mineral compositions. Observed from optical to radio wavelengths, each band reveals unique information about the dust populations. Optical and near-infrared observations are specifically sensitive to light scattered off the surfaces of the micron-sized particles. Here, we present results from the JWST/NIRCam GTO program 2780, designed to observe five disk systems previously identified to be exceptionally bright at optical wavelengths (HD 10647, HD 32297, HD 61005, HD 107146, and HD 181327) with six filters using the NIRCam coronagraphs. The NIRCam data complement previous shorter-wavelength images of these same systems. They reveal scattered light from the disks and from the extended halos of tiny grains under the influence of radiative forces, at high resolution and signal to noise. All the systems show evidence for water ice, although it can have differing radial distributions and tends to show stronger signatures in the halos. In the two cases we could analyze, the scattering phase function in the disks resembles the behavior of dust in the Solar System with evidence for enhanced forward scattering in the halos, consistent with the latter being composed of tiny grains. MIRI images for two systems are more centrally concentrated than the shorter wavelength ones, suggesting a role for dragged-in larger grains.
Read the full abstract
Adaptation and Self-Organizing Systems
Low-Dimensional Reduction Theory for Populations of Phase Oscillators with a Gaussian Frequency Distribution
Low-dimensional reduction theories such as the Ott-Antonsen ansatz have played a crucial role in the study of populations of coupled oscillators. Their application, however, has largely been restricted to systems with frequency distributions of rational-function form, such as the Cauchy distribution. For such distributions, the residue theorem allows the dynamics of the global order parameters to be closed in terms of a finite number of poles, thereby yielding a finite-dimensional system of ordinary differential equations. Rational frequency distributions, however, generally have heavy tails and only finitely many well-defined moments, and therefore may not always be realistic as frequency distributions. In this paper, we develop an approximate low-dimensional reduction theory based on perturbation theory for weakly heterogeneous populations of phase oscillators with a Gaussian frequency distribution. We construct the theory not only for first-harmonic coupling, for which the Ott-Antonsen ansatz applies, but also for populations of phase oscillators with multi-harmonic coupling. The effectiveness of the proposed low-dimensional reductions is demonstrated through both theoretical analysis and numerical simulations.
Read the full abstract
Rings and Algebras
Free Novikov-Zinbiel algebra
Let $A$ be a commutative-associative algebra with an invertible derivation $D$, and put $R=D^{-1}$. We study the operations \begin{equation*} x\succ y=R(x)y,\qquad x\prec y=xD(y), \end{equation*} which define a Novikov-Zinbiel algebra. Using a simple graded model, we represent multilinear $\prec,\succ$-monomials by rational functions and construct an explicit basis for the resulting space. This yields a description of the multilinear components of the free Novikov-Zinbiel algebra. As a consequence, we prove that every multilinear identity satisfied by this construction follows from the defining identities of Novikov-Zinbiel algebras.
Read the full abstract
Algebraic Geometry
Disjoint and nearly disjoint sums of matrix multiplication tensors and their centroids
This paper addresses centroids, which are fundamental invariants of tensors. Our main results are as follows: (i) The construction of explicit tensors with very large centroids, whereas previously it had been conjectured that none such exist. (ii) An upper bound on the dimension of the centroid that is essentially attained by our examples. (iii) The development of a geometric technique to write down border rank decomposition of tensors using centroids and "extended centroids". (iv) The technique is applied to tensors of this paper to prove they are of minimal border rank. The technique is versatile and enables us to geometrically derive and improve upon previous ad hoc decompositions. (v) The construction of symmetric tensors with large centroids and proof that they are wild in the sense of Buczyńska-Buczyński. Our results also pave the way for new upper bounds on the exponent of matrix multiplication. The geometric technique also constructs new "better" tensors for Strassen's laser method from old, and we apply this to the tensors of Strassen and Schönhage to get better tensors in the sense that they give better upper bounds on the exponent than the original tensors.
Read the full abstract
Differential Geometry
Scalar curvature on Kähler blow-ups and systolic inequalities
In this paper, we develop the weighted level set method for a Kähler manifold $(X^n,\omega)$ admitting an almost holomorphic map to a possibly singular base $Z$, which is not uniruled. As a key intermediate result, we prove that any blowup $\operatorname{Bl}_SX$ of $X$ along smooth submanifolds $S$ of $ \operatorname{codim} S\ge2$ admits a sequence of Kähler metrics with scalar curvature globally and arbitrarily $C^0$-close to the scalar curvature of $\omega$. As a consequence, we establish the sharp \(2\)-systole estimate for every positive scalar curvature Kähler manifold $(X,\omega)$ and prove $\min_XS(\omega) \cdot\operatorname{sys}_2(\omega) \le 2\pi r(r+1)$, where \(r\) is the rational dimension of $X$, with equality if and only if the universal cover splits as $(\widetilde X,\widetilde \omega) \cong (\mathbb P^r,\omega_{\mathrm{FS}}) \times(Y^{n-r},\omega_{\mathrm{RF}})$ up to normalization where $\omega_{\mathrm{FS}}$ is the Fubini-Study metric and $\omega_{\mathrm{RF}}$ is Ricci-flat. We also show a sharp even-systolic inequality in the same setting when the general fibre is the projective space.
Read the full abstract
Combinatorics
Closing the gap and settling the problem of queens on an $n\times n$ board, each attacking at most one other
Let $q(n)$ denote the largest number of queens that can be placed on an $n\times n$ chessboard so that no queen attacks more than one other queen. We prove that $q(n)=\lfloor4n/3\rfloor$ for every $n\geqslant6$, and that $q(n)=n$ for $n\leqslant5$, which settles a previously conjectural value. As a corollary, we also settle that, in the version of the problem where each queen attacks exactly one other queen, the answer is $2\lfloor2n/3\rfloor$, again as previously conjectured.
Read the full abstract
Analysis of PDEs
Monotonicity of the propagation speed with respect to the diffusion in a Lotka-Volterra competition-diffusion system
In this paper, we study the dependence of the propagation speed on the diffusion ratio in a Lotka-volterra competition-diffusion system under strong competition. The model is known to admits a unique monotone travelling front connecting two exclusion equilibria and the sign of its speed determines which species invades the territory occupied by the other. We establish smoothness results for the wave speed and the wave profile. Our main result shows that the wave speed is a strictly decreasing function of the diffusion ratio on explicit unbounded regions of parameter space. This monotonicity had been observed numerically but, had not previously been proved. The proof combines the smooth dependence of both the travelling front and its speed on the diffusion ratio, an adjoint identity for the linearized operator and a maximum principle argument. At zero speed, we also prove that there exists a smooth threshold, we obtain an exact formula for its derivative and deduce its monotonicity. These threshold results also provide a complete characterization of the sign of the wave speed.
Read the full abstract
Quantum Physics
Factorized Boolean representations for efficient quantum synthesis
Quantum algorithms promise advantages beyond classical reach, but running them on error-corrected hardware requires translating Boolean specifications into reversible circuits, and the resources that translation demands determine what is executable. Established methods minimize a Boolean expression and map it to a circuit, assuming the minimized form is best. Here we show that minimized expressions retain algebraic structure minimization cannot reach, arising from containment and complementary-polarity relationships among their terms, and that extracting it yields circuits cheaper to execute despite having more operations. The decisive quantity is not a circuit's operation count but the control count of its widest operation, a superlinear cost; extracting shared factors trades a few wide operations for many narrow ones and reduces qubit count. Across benchmarks and oracles from quantum search and factoring algorithms, at the representation level the transformation never increases either cost measure, a guarantee from its construction. Translation to an executable circuit returns part of that advantage, since auxiliary lines must be uncomputed, yet the factorized circuit still left a leading circuit-level optimizer reaching lower final counts, and faster, than unaided. The representation of a computation is therefore itself a resource, optimizable before compilation and distinct from both logic minimization and circuit-level optimization.
Read the full abstract
Artificial Intelligence
Mechanistic Reaction Prediction via Discrete Flow Matching on Graph-Structured Electron Occupation
Chemical reactions are fundamentally transformations in electron space, yet most machine learning approaches model them either through de novo generation of product molecules or through heuristic graph edits that operate directly on molecular topology. We introduce MAELLE (MechAnistic Edit fLow-matching on eLectron rEarrangements), which instead models reactions as discrete flow matching over electron occupation vectors. Concretely, we formulate the reactant-to-product mapping as a Continuous-time Markov Chain (CTMC) over the graph-structured integer-valued electron occupation space defined on all bonding, non-bonding, and hydrogen sites. To construct the interpolants between the reactants and products, we generalize the discrete flow matching mixture path to an edit-based formulation, where the electron moves are interpolated using Optimal Transport, yielding a mechanism-like set of moves without elementary step annotations. MAELLE achieves competitive performance on the USPTO-480K benchmark compared with leading reaction prediction models. Beyond in-distribution learning, we evaluate robustness across two out-of-distribution settings - structural complexity and reaction type - and find that MAELLE maintains strong performance where existing methods degrade. Finally, because the learned flow operates over the full electron redistribution, MAELLE naturally recovers mechanistic trajectories that align with known chemistry and can predict side products of a reaction.
Read the full abstract
Computation and Language
Stochastic Estimation of Transduced Language Models
Transduced language models (TLMs) compose a pretrained source language model with a functional finite-state transducer to induce a language model over target strings. Computing the probability of a target prefix under a TLM amounts to summing the source-model probabilities of all source strings that the transducer maps to target strings beginning with that prefix. This set can be exponentially large or infinite. Prior work uses a computational shortcut based on source prefix probabilities, then approximates the resulting sum with threshold-pruned beam summing. This produces a lower bound with unknown error. Instead, we resample source prefixes without replacement and reweight each selected prefix by the inverse of its inclusion probability. We show that applying this correction recursively gives an unbiased estimator of the target prefix probability and lets us estimate the mass lost by threshold pruning. Our beam-summing algorithm extends the retained source prefixes and samples which prefixes to keep, reducing their number as more probability mass is added to the running estimate. This can save computation and guarantees that the run halts with probability one. We evaluate the method on encyclopedic text and DNA against sequential Monte Carlo baselines that resample with replacement. It achieves a better compute–variance tradeoff on text and lower error at the same maximum number of particles on DNA. On a DNA-to-amino-acid transduction, it reduces runtime by several orders of magnitude relative to threshold-pruned beam summing and makes estimating prefix probabilities for long target strings feasible. Replacing threshold pruning with unbiased sampling in a published reading-time analysis substantially lowers the estimated corpus surprisal but leaves the published conclusions unchanged.
Read the full abstract
Software Engineering
Persona-Execution Separation: An Architecture Pattern for Evolving LLM Agents under Execution Audit
Large language model (LLM) agents in governed organizations must let the persona (instructions, tone, self-presentation) evolve freely, while keeping execution (stateful, audited work) traceable. A single trust domain does not satisfy both cheaply. We present Persona-Execution Separation (PES): persona and execution reside in different trust domains, connected by a governed contract bridge. The persona is singly-homed and may drift; execution is faceless and audited. Status summaries may return; data bodies remain in the restrictive domain except a data-loss-prevention (DLP) exception; identity stays continuous. An approval matrix, DLP, and audit enforce the crossing. PES follows from three goals: free drift, execution traceability, and decoupling. Under LLM representational indistinguishability, any single-domain mechanism meeting all three must re-introduce typed change objects, an external gate, and a stable audit anchor: PES rebuilt at higher coupling cost. A development/pilot case in a regulated platform records five decisions over one month, four with rejected alternatives. A mechanism check found no execution-side re-validation under persona perturbation (five configurations) and no persona fingerprint on hard-asserted fields of completed runs. A controlled replication in regulated coding agents reproduced the separation under isolation across five models and four providers; bridge overhead was under 0.2% of end-to-end time in both environments. A probe of a pre-separation build found the execution path decoupled from the persona by omission, not by construction. The pattern applies when multi-user deployment, execution audit, and persona churn hold jointly.
Read the full abstract
Quantum Physics
Detuning- and Stark-robust Rydberg gates
Rydberg entangling gates driven by a two-photon transition in alkali atoms suffer from an adverse scaling of light-shift-induced errors. Robustness to such detuning errors is known to be impossible to achieve in the design of conventional Rydberg gate protocols, where only one of the qubit states is coupled to the Rydberg state. Here, we show that in a more general framework, in which both qubit states take part in the gate, full or partial robustness to these errors can be realized. We present two gate constructions, which either cancel the errors outright or convert them into single-qubit errors that can be corrected locally. We map the regimes – in terms of light-shift strength, intensity inhomogeneity, and Rydberg decay rate – in which these protocols outperform the widely used time-optimal Rydberg gate, and find that they already include the conditions of state-of-the-art experiments. Finally, we show the existence of a Rydberg `fly-by' entangling gate, an important primitive for an emerging class of neutral-atom quantum computing architectures.
Read the full abstract
Combinatorics
Leading term strandings for webs
A web is a plane graph encoding an invariant vector in a tensor product of fundamental representations of a quantum group. A stranding of an $\mathfrak{sl}_n$ web is a system of colored oriented curves recording one monomial of the vector it encodes. This article focuses on identifying and constructing leading term strandings, those recording the leading term of a web's vector with respect to a lexicographic order on monomials. We show that every open strand of a leading term stranding is clockwise, so the boundary data of such a stranding is recorded by a row-strict tableau. From a row-strict tableau, we construct a web with a prescribed leading term, and the resulting webs form a web basis. This non recursive construction recovers Fontaine's bases and produces others. For $\mathfrak{sl}_3$ webs with no flat vertices, we identify a leading term stranding using the depths of the faces of the web. Finally, we ask whether any two strandings of a web are connected by a sequence of local moves called strand reversals. We answer this question affirmatively for $\mathfrak{sl}_3$, extending a known result for closed webs to webs with boundary; in particular, a leading term stranding of any $\mathfrak{sl}_3$ web is always reachable from an arbitrary stranding.
Read the full abstract
Cryptography and Security
Beyond F1: Evaluating Coverage and Failure Recovery in AI Model Security Scanners
Static scanners are increasingly used to identify executable or otherwise unsafe content in machine- learning artifacts, yet conventional evaluation metrics characterize only cases where a scanner yields a usable security judgment. We evaluate ModelScan, ModelAudit, and Fickling using a controlled, artifact-backed benchmark on a synthetic corpus of 170 Pickle and PyTorch focused artifacts across 145 specimen families, 135 of which have binary security ground truth and 10 of which are intentionally malformed without labels. We explicitly distinguish non-N/A coverage, analysis completion, definitive security decisions, non-security findings, and unsupported outcomes. On labeled families, ModelAudit produced definitive security decisions for all 135 families (100%), Fickling for 110 (81.5%), and ModelScan for 67 (49.6%). Conditional on making a definitive judgment, ModelScan achieved 100% precision, recall, and F1. Fickling identified no unique true- positive families beyond those found by the combination of ModelAudit and ModelScan. Furthermore, for the 48 malicious families where ModelScan failed to complete its analysis, both ModelAudit and Fickling generated detections consistent with ground truth. These findings underscore the need to separate judgment accuracy from judgment availability, as well as incremental detection coverage from tool-level redundancy.
Read the full abstract
Fluid Dynamics
High-order stabilized matrix-free simulation of rotating mixing devices using the Mortar Element Method
We present a finite element framework to simulate rotating mixing devices using the Mortar Element Method as a domain decomposition strategy. The model is implemented within a matrix-free Navier-Stokes framework which uses a high-order Continuous Galerkin method. The discretized domain is subdivided into rotor and stator parts. An Arbitrary Lagrangian-Eulerian approach accounts for the relative rotor-stator motion, and stabilization is ensured through the Streamline-Upwind/Petrov-Galerkin and Pressure-Stabilizing Petrov-Galerkin methods. The rotor-stator domains are connected by an interface composed of mortar cells, and continuity is weakly enforced in a Discontinuous Galerkin fashion by accounting for boundary integrals at the rotor-stator interface. Verifications of the convergence order in two-dimensional steady and transient examples report optimal rates. The geometric non-conformity created at the mortar interface due to rotor rotation does not introduce significant error in the solution. A three-dimensional example is used to investigate the model's scalability, which yields ideal strong scaling for large problems. A two-dimensional Rushton impeller example uses a torque analysis to showcase the mesh convergence, and the corresponding velocity profile is in agreement with existing numerical results. In a three-dimensional pitched blade turbine case, the power number curve (Np vs Re) shows good agreement with experimental data for Reynolds number values from 1 to 2000. An energy balance analysis reports a numerical dissipation of 1% for Re=200 and of 10% for Re=2000. By exploiting modern hardware capabilities through matrix-free methods, the proposed model is a robust, accurate, and efficient framework suitable for simulating flows with rotating geometries.
Read the full abstract
Information Retrieval
misi: a Metric Inverted Sample Index
We present misi, an inverted index for approximate nearest-neighbor search over general metric spaces whose vocabulary is a random sample of the database, of size proportional to $n$. Each object is represented by its $k_b$ nearest sample points, found by a pluggable inner index over the sample; queries are answered by an idf-weighted shared-neighbor vote followed by exact verification of $C$ candidates. The construction generalizes the NAPP index from a constant number of pivots to a linear-size vocabulary, which keeps posting lists at constant expected length $\rho = k_b/\alpha$ as $n$ grows and turns the index into a combinator: any high-recall index on $\alpha n$ points yields an index on $n$ points, for any metric. A probabilistic model gives a recall guarantee – $k_b$ logarithmic in $n$ over the overlap gap suffices, with a verification budget the index itself estimates – and a matching limit: the vote cannot resolve overlap differences below order $1/\sqrt{k_b}$. The design's strengths are structural: construction is $n$ independent searches – embarrassingly parallel, deterministic, $5{,}250$ s for $10^8$ vectors on 64 cores, $3.7\times$ faster than a matched-recall graph build – it streams under an enforced 3 GiB cap, and the portable artifact serves $10^8$ vectors from NVMe within an enforced 8 GB budget, below the working floor of the SSD-graph baseline. Its cost is query-time work: saturated graph baselines answer $6$-$16\times$ faster in RAM, and the verification budget for 0.99 recall grows as $n^{0.30}$. All results carry seeds, saturation sweeps and full configurations, are generated from run manifests, and include measured negative results. The intended applications weight construction cost, determinism, memory footprint, or black-box metrics over peak throughput: frequently rebuilt corpora, batch similarity workloads, constrained-memory serving.
Read the full abstract
Artificial Intelligence
Learning a Continuous Sepsis Severity Score Without Hour-by-Hour Supervision: A Two-Site Retrospective Study
Currently used sepsis severity indices rely on fixed variables and weights established decades ago, which are coarsely discretized and calibrated to a cohort that no longer reflects contemporary critical care. No alternative learned directly from patient trajectories is in routine use. We conducted a retrospective two-cohort study on a total of 29,116 and 7,691 adult patients meeting Sepsis-3 criteria from two hospital systems in Massachusetts and Georgie, respectively. We developed a sepsis index using 43 routinely charted variables over a 72-hour treatment window. Unlike previous studies, we use mortality as a treatment-level ranking signal rather than a per-state target, allowing credit to be redistributed non-uniformly across timesteps. Evaluation was done on a permanent 20% test holdout, using clinical vignettes and Spearman correlation. Uncertainty intervals were obtained by bootstrap resampling of whole patients. Under this ranking scheme, non-survivors scored 1.19-1.64 points higher than survivors on a 0-10 scale within all strata of baseline SOFA-2, with similar results stratifying within lactate, mean arterial pressure (MAP), and creatinine. Within-patient change in the index correlated with change in lactate (Spearman rho = 0.39; n = 1,854). Similar, weaker correlations were found for MAP and creatinine. On a cohort level, cross-institutional agreement measured by Spearman correlation between models trained on different sites, were 70-77% of same-site correlation. External within-patient correlations were 0.54 and 0.59 against ceilings of 0.92 and 0.90. Our index also correlated with established indices, while null controls stayed near zero. Our index demonstrated hourly prognostic information that meaningfully separates patient outcomes and is consistent with clinical expectation, indicating potential as a decision support tool complementing clinical judgement.
Read the full abstract
Computation and Language
Boosting LLM Exploration via Weak-Model Guidance in RLVR
Reinforcement Learning with Verifiable Rewards (RLVR) significantly improves LLM reasoning but often causes a drop in policy entropy, leading to narrowed reasoning coverage and degraded pass@$k$ for large $k$. While existing methods mitigate this entropy collapse through algorithmic regularizations, cross-model non-parametric perturbation is also neglected. In this work, we propose a simple yet effective approach to preserve the generative diversity of LLMs during RLVR. Instead of relying solely on internal exploration, we force the target model to generate answers based on partial reasoning trajectories generated by a smaller, weaker language models. These unfamiliar prefixes effectively disrupt over-confidence and encourage the exploration of distinct reasoning paths. We empirically study the potential of outer prefixes, revealing the mechanism of the impact of distributional discrepancy to the exploration dynamics in RLVR training. Experiments across multiple mathematical benchmarks show that our method consistently outperforms vanilla RLVR. Notably, the performance gain becomes increasingly pronounced as $k$ scales up, demonstrating a substantial expansion of reasoning coverage. Furthermore, our approach efficiently mitigates entropy collapse without requiring additional SFT, intricate reward designs, or complex prompting.
Read the full abstract
Game Theory
Algorithms for Robbins' Problem using Markov Decision Processes
In this paper, we consider Robbins' problem, which is a full information variant of the well-known secretary selection problem. In this version of the problem, the goal is to minimize the expected rank of the selected candidate among $n$ that are interviewed sequentially, and a decision to select or not the $m^{th}$ candidate needs to be taken right after the interview (so without seeing the last $n-m$ candidates and without recall). We first show how to model instances of Robbins' problem as infinite Markov Decision Processes (MDPs). Then we propose several finite-state abstractions of these MDPs that allow us to approximate the value of the problem for fixed $n$. While it is known that the full memory of past candidates' values is necessary for optimal expected rank minimization, making the analysis of the problem challenging, we highlight simple memory structures that are sufficient for obtaining near-optimal selection strategies. Additionally, we provide approximate values for Robbins' problem for numbers of candidates $n$ up to 100 for which no good approximations were previously known (the exact value is only known for instances where $n \leq 4$ and numerical approximations were for small values of $n$ not exceeding one digit), for all $n : 5 \leq n \leq 100$, we give better approximation than what was previously known.