Peer Revenue Equity in science

Papers

Fresh from arXiv, typeset properly. Highlighted words trade on the market.

All of arXiv

Thursday 27 August

Computation and Language

Between the Lines: Multi-turn Contextual Investment Advice Detection

As large language models (LLMs) increasingly power financial conversational assistants, detecting investment advice in natural language has become a critical regulatory and safety challenge. Unlike explicit single-turn recommendations, investment advice often emerges implicitly and progressively over multi-turn conversations, shaped by user intent, delivery style, and contextual buildup. We introduce the first comprehensive framework for detecting both sequential and contextual investment advice in dialogue settings. Central to our approach is a modular, top-down synthetic data generation pipeline that systematically spans a taxonomy of advisory trajectories using an axis-bucket-descriptor schema. We simulate both benign and high-risk conversations across diverse user personas, financial instruments, and risk tiers, and evaluate detection capabilities using LLM-as-a-Judge with entropy-based uncertainty diagnostics. Our results show that contextual advisory detection remains substantially more challenging than sequential cases, even under few-shot prompting with tailored demonstrations. This work lays the foundation for scalable guardrail development in high-stakes, compliance-sensitive domains, where advisory risk is often subtle, distributed, and perception-dependent.
Read the full abstract

All authors Mingshu Li, Adam Boncz, Mate Toth, Justin Williams, Justin Chan, Dhagash Mehta

Information Retrieval

Explain, Then Forget: Recommendation Unlearning under Asymmetric Access Constraint

Recommendation unlearning aims to remove the influence of specific historical interactions that should no longer affect future recommendations, such as noisy clicks or outdated user preferences. However, performing recommendation unlearning becomes much harder in modern Machine-Learning-as-a-Service (MLaaS) architectures, where the recommendation model is remotely hosted by a model provider (MP) and accessed by service operators (SO) only through APIs. In this setting, a service operator serves end users but cannot access or modify internal model parameters. As a result, existing unlearning methods are difficult to apply because they require access to model internals for retraining, fine-tuning, or other parameter updates. To address these challenges, we propose Causal Explanation-based Unlearning (CEU), a novel unlearning framework tailored for MLaaS environments. Instead of modifying internal model weights, CEU operates through a two-step output-level score intervention. First, it estimates how the forgotten interaction causally contributes to the current recommendation. Second, guided by this causal signal, the system removes the influence of the forgotten interaction by adjusting the final ranking scores of affected items. Because CEU leaves model parameters untouched, it naturally supports continuous unlearning requests while avoiding the error accumulation that can result from repeated parameter updates. Extensive experiments on three public datasets show that CEU achieves unlearning effectiveness comparable to full retraining while preserving recommendation utility and significantly reducing computational cost. The source code is available at: https://github.com/zhang-haichao/Explain-and-Forget.
Read the full abstract

All authors Haichao Zhang, Chong Zhang, Wen Wang, Shi Qiu, Peiyu Hu, Jia Wang

Functional Analysis

A functional model for hyponormal contractions

In this paper, we establish an analytical model for a class of hyponormal contractions $T$ satisfying $Ran[T^*,T] \subseteq \mathcal D_{T^*}\ominus \mathcal D_T$, where $\mathcal D_{T^*}$ and $\mathcal D_T$ are the associated defect spaces for $T$. We also show that, under a certain condition, a contraction $T$ with $\mathcal D_T \subseteq \mathcal D_{T^*}$ is hyponormal.
Read the full abstract

Information Retrieval

Controllable Generative Recommendation via Guided Token Refinement

Generative recommendation over discrete Semantic IDs has become a powerful paradigm for item retrieval. Real-world systems, however, frequently require real-time controllability to balance accuracy against objectives such as diversity. Existing autoregressive methods generate tokens strictly from left to right. This fixed direction locks in early decisions, preventing holistic planning and making control difficult. Consequently, direct interventions often break the semantic structure, producing non-existent items. To address this, we propose ReGen, a framework that shifts from sequential decoding to parallel iterative refinement. ReGen reformulates generation as a mask-filling task, predicting all tokens simultaneously rather than one by one. This global view allows us to apply control signals directly to tokens with high-level semantics without retraining the model. To ensure that decoded candidates map to real items, we introduce a topology-constrained tree that iteratively corrects low-confidence predictions while strictly enforcing valid codebook paths during beam-based inference. Experiments show that ReGen provides a favorable accuracy-diversity trade-off under moderate guidance while maintaining high validity of generated Semantic IDs. The source code is available at: https://github.com/zhang-haichao/ReGen.
Read the full abstract

All authors Haichao Zhang, Zhixuan Liang, Chong Zhang, Zixi Chen, Wen Wang, Jia Wang

Functional Analysis

Weighted Two-Parameter Quantum Divergences Generated by Matrix Power-Mean Gaps

Let $\mathbb{P}_n$ be the cone of $n\times n$ complex positive definite matrices. For $0<\theta<1$ and $r\ge0$, let $M_r^{(\theta)}$ denote the weighted matrix power mean, with the logarithmic endpoint at $r=0$, defined by $$ M_r^{(\theta)}(A,B) = \begin{cases} \big((1-\theta)A^r+\theta B^r\big)^{1/r}, & r>0,\\[2mm] \exp\big((1-\theta)\log A+\theta\log B\big), & r=0. \end{cases} $$ For $p>q\ge0$, we introduce $$ \Phi_{p,q}^{(\theta)}(A,B) = \operatorname{Tr}\big[M_p^{(\theta)}(A,B)-M_q^{(\theta)}(A,B)\big]. $$ We prove that $\Phi_{p,q}^{(\theta)}$ is a quantum divergence in the sense of Bhatia-Gaubert-Jain. Moreover, for $$ 0\le q\le1\le p\le2,\qquad p>q, $$ it is jointly convex and satisfies the data processing inequality for completely positive trace-preserving maps, with the output side interpreted by the regularized boundary convention whenever it is singular. The construction recovers Hellinger-type and Jensen-Shannon-type quantities as special cases or limits. We also compute the induced local quadratic form and provide scalar and matrix obstructions showing that the convexity and data-processing range cannot be extended to all parameters.
Read the full abstract

All authors Trung Hoa Dinh, Minh Toan Ho, Ha Trang Nguyen, Trung Dung Vuong

Instrumentation and Detectors

The efficiency of moderating neutron detector: Monte Carlo simulation, experimental validation, and angular dependence

This paper describes the Monte Carlo simulation and experimental validation of the neutron detection efficiency in a 4pi polyethylene moderating detector. The experimental validation was performed using several sources providing neutron energies from 10s of keV up to 14 MeV. The Monte Carlo simulations employed the MCNP6 code are believed to be accurate within 3% for neutron energies below 6 MeV. A new method for parametrizing the dependence of the efficiency on the neutron angular distribution for in-beam experiments is also presented. This approach utilizes a Legendre decomposition of the efficiency and allows all of the needed information to be extracted from a single simulation.
Read the full abstract

Combinatorics

The Eigenvalue Region of a Family of Markov Chains with Randomness Confined to Two States

We determine the complete eigenvalue region of the following Markov chains on $2q$ states, for $q\ge2$. The states are $A_0,\ldots,A_{q-1}$ and $B_0,\ldots,B_{q-1}$. Each transition $A_j\to A_{j+1}$ and $B_j\to B_{j+1}$, for $0\le j<q-1$, has probability one. From $A_{q-1}$, the chain moves to $A_0$ with probability $a$ and to $B_0$ with probability $1-a$. From $B_{q-1}$, it moves to $B_0$ with probability $b$ and to $A_j$ with probability $(1-b)p_j$. Here $a,b\in[0,1]$ and $p_j\ge0$ with $\sum_{j=0}^{q-1}p_j=1$; these parameters vary over all permitted values. Every segment from zero to an attainable eigenvalue lies in the region. Its unit-circle points are exactly the roots of unity of order at most $2q$, as for unrestricted stochastic matrices of the same order, yet the full region is strictly smaller. It excludes $0<|\operatorname{Arg}\lambda|<\pi/q$. For odd $q$, the final nonreal boundary arc ends at a negative real point inside the unit disk; the remaining interval to $-1$ is also boundary. Every transition matrix has the same characteristic polynomial as one with $a=b$. Among $q+1$ fixed matrices in this balanced family, convex combinations of at most three realize every attainable eigenvalue, and two suffice on the boundary. Both bounds are sharp within this set. Consecutive reduced fractions in $[1/(2q),1/2]$ with denominators at most $2q$ select the upper nonreal boundary constructions. Each boundary radius is the unique solution of an equation whose left-hand side increases strictly with the radius.
Read the full abstract

Machine Learning

Active Client Selection in Federated Trajectory Prediction with Uncertainty-Awareness and Heterogeneous Complexity

Training sequence models such as transformers is now standard for autonomous vehicle trajectory prediction, yet assembling high-quality centralized datasets remains challenging because real-world trajectories are fragmented across regions and vehicles. Federated Learning (FL) offers a natural alternative, but faces two distinctive challenges: high scene uncertainty arising from trajectory or map ambiguity, and cross-scene complexity heterogeneity caused by diverse map topology, traffic density, agent composition, and driving behaviors. We propose a family of active client selection methods that progressively incorporate awareness of scene uncertainty and complexity to prioritize informative clients. Our uncertainty-aware selectors use per-client negative log-likelihood under an uncertainty-aware global objective and estimated aleatoric uncertainty. We further develop a selector that jointly considers scene complexity and uncertainty, motivated by the intuition that knowledge from complex scenes can transfer to easier ones. Experiments on Argoverse show that federated trajectory prediction outperforms locally trained models. Uncertainty-aware selection accelerates convergence and improves minADE, minFDE, and MR. Under strong scene-complexity heterogeneity, our joint complexity- and uncertainty-aware selector achieves the best generalization and further accelerates convergence, demonstrating the benefit of prioritizing complex and informative scenes.
Read the full abstract

All authors Yiming Xie, Muzi Peng, Fei Miao, Ningfang Mi, Lili Su

Artificial Intelligence

PEEL: Physics-Enabled Evidential Learning for Identifiable Uncertainty in CT Imaging

Normal-inverse-gamma (NIG) regression is not uniquely identifiable from its marginal Student-t likelihood: the likelihood determines three combinations of four NIG parameters and is constant along a one-dimensional fiber. We identify that fiber using independent physical measurement. As an initial embodiment, a reconstruction network receives one noisy filtered-backprojection (FBP) image and is first trained only by Student-t negative log-likelihood to estimate the three identifiable coordinates (gamma, alpha, c). The network is then frozen; repeated physical-noise realizations propagated through its reconstruction output form a Monte Carlo (MC) teacher label for output-domain aleatoric variance. An aleatoric head attached to frozen features learns this label, after which (beta, nu) are recovered algebraically. On 30 held-out simulated objects at five photon levels, one-image predictions achieved pooled Spearman correlations of 0.832-0.951 against independent 400-repeat references, median within-image correlations were 0.834-0.947, and 98.81-99.55% of evaluated pixels satisfied the algebraic admissibility condition. The method needs no KL term, reference prior, evidence regularizer, or cross-loss weight.
Read the full abstract

Algebraic Geometry

Cartan Determinants in algebraic geometry

We formulate a dg-categorical analogue of the Cartan determinant conjecture using numerical Grothendieck groups. Let $D$ be a smooth proper dg category over an algebraically closed field $k$. The Euler pairing on $K_0(D)$ descends to a nondegenerate bilinear form on the numerical Grothendieck group $K_{\mathrm{num}}(D)$, and hence defines a numerical Cartan determinant $\det M_\chi(D)$. While the classical Cartan determinant conjecture predicts that $\det M_\chi(A)=1$ for finite-dimensional smooth algebras $A$, this equality is no longer expected for general smooth proper dg categories. However, we propose the following positivity question: $\det M_\chi(D)>0.$ Our main result proves this positivity when $D=\operatorname{Perf}(X)$ for a large class of smooth projective varieties $X$. More precisely, we show that $\det M_\chi(X)>0$ if either $\dim X$ is odd, or $X$ has even dimension and satisfies the numerical Hodge standard conjecture.
Read the full abstract

Other Statistics

How many cards, until the first ace: variations, extensions, lachrymae, confidence, Dirichlets

From a deck of cards, how many cards do I need to draw, until the first ace? I identify the distribution for this waiting time $T$, and its satisfyingly nice expected value ${\rm E}\,T=(N+1)/(n+1)$, with $N$ the number of cards and $n$ the number of aces; hence $53/5=10.6$ for the standard setup. After having solved this Question One I go on to certain alternative solutions and extensions, involving e.g. Beta approximations. I also consider the distributions and means for the 2nd, the 3rd, the 4th occurrences of aces, with generalisations, where there is a Dirichlet distribution in wait for us, with further links to order statistics for the uniform. Furthermore, an apparatus is developed for obtaining estimators and full confidence distributions for applications where one knows the number $n$ of aces, but not the deck size $N$; and correspondingly for inference about the unknown population size $N$ when $n$ is known. If you have 1000 people in a room, and need to interview 11 of them until you've found the first left-handed person, how may left-handed are there in the room – here we need both an estimate and a clear measure of uncertainty.
Read the full abstract

Materials Science

Building Atomic Structures from Natural Language

Large language models can write the input parameters of a materials simulation from a prompt, but they cannot build the atomic structure those parameters describe. A crystal, a surface slab or a mismatched interface must satisfy global constraints-periodicity along specific axes, a given vacuum thickness, a commensurate registry, sensible bond lengths. Generation one token at a time enforces none of them, and a violation is invisible in the output. We present ATLAS, a framework that makes structural construction scriptable. Seven components-analyzer, LLM skill, JSON format, build engine, validator, scorer and web interface-are organized around a component algebra of named structures and spaces. Every specification carries a check list, so a build is verified against the request rather than merely parsed from it, and builds are scored against the user's intent. Requests state physical sizes, and the engine computes the replication that reaches them. The translation is the one stochastic stage, and it is recorded, so a run replays with the model held fixed-which is what makes the framework's own behaviour measurable rather than merely observed. Fourteen structural classes are demonstrated, from bulk cells and nanoparticles through surfaces, defects and interfaces to amorphous networks and liquids. We outline how the same loop generates training sets for machine-learning force fields.
Read the full abstract

Number Theory

Reducibility of $rx^m+p^e f(x)$ for Large Primes $p$

We study the reducibility over $\mathbb{Q}$ of $F_{p,e}(x)=rx^m+p^ef(x)$, where $r\in\mathbb{Z}\setminus\{0\}$, $f\in\mathbb{Z}[x]$, and $0\le m<n:=\mathrm{deg}\, f$, for primes $p$ above explicit coefficient-dependent thresholds. For arbitrary $e\ge1$, we determine the degrees, endpoint $p$-adic valuations, reductions modulo $p$, and heights of all nonconstant proper integral factors. Writing $\delta=\gcd(e,m,n)$, we show that every such factor has degree $tn/\delta$ for some $1\le t\le\delta-1$; in particular, $\delta=1$ implies irreducibility. For $e=2$ and $e=3$, we obtain effective necessary and sufficient criteria in terms of at most two associated binary quadratic forms and at most one associated binary cubic form, respectively. The quadratic criterion reduces to at most two fixed Pell-type equations with prescribed second coordinate $p$. In the cubic case, Thue's theorem yields finiteness of the reducible primes whenever the associated form is irreducible over $\mathbb{Q}$.
Read the full abstract

Number Theory

The Fermat-Type Matrix Equation over $\mathrm{GL}_2(\mathbb{Z})$

We determine the complete ordered solution set of the Fermat-type matrix equation $X^n+Y^n=Z^n$ in $\mathrm{GL}_2(\mathbb{Z})$ for every integer $n\ge3$, expressed in terms of simultaneous integral-conjugacy orbits while keeping the variable symmetries of the equation separate. The equation is solvable if and only if $4\nmid n$ and $6\nmid n$. Thus, in contrast with the determinant-one setting, odd multiples of $3$ are solvable in the full group $\mathrm{GL}_2(\mathbb{Z})$ and necessarily involve determinant-$-1$ components. For every admissible even exponent, the full ordered solution set has a canonical disjoint orbit decomposition indexed by the same transversal that occurs in the matrix Pythagorean equation. For odd exponents with $3\nmid n$, the commuting solutions form exactly six simultaneous-conjugacy orbits. The noncommuting odd solutions form exactly twenty simultaneous-conjugacy orbits when $3\nmid n$, and exactly fourteen when $3\mid n$. Under the coarser equivalence generated by signed variable permutations and simultaneous integral conjugation, these noncommuting orbits collapse to four and three classes, respectively. The proof combines the canonical orbit classification of the matrix Pythagorean equation, Cayley–Hamilton recurrences, commutator divisibility, trace-growth estimates, integral conjugacy of torsion matrices, and an elementary symmetry-reduced trace argument for the cubic equation.
Read the full abstract

High Energy Physics (Theory)

Unification of Einstein Gravity with Internal Interactions

The gauge-theoretic formulation of gravity, enriched with the observation that the tangent group of a curved manifold need not have the same dimension as the manifold itself, provides a natural framework for unifying gravity with internal interactions. In the present work we study the unification of Einstein gravity with the $SO(10)$ grand unified theory within an $SO(1,17)$ gauge theory. As a central result, we derive Einstein gravity with a cosmological constant from the spontaneous symmetry breaking (SSB) of an $SO(1,5)$ gauge theory to the Lorentz group via two independent routes: one using two scalar fields in the fundamental representation $\mathbf{6}$ of $SO(6)$, and one using a single scalar in the adjoint $\mathbf{15}$. Both yield a Gauss-Bonnet + Einstein-Hilbert + cosmological constant action. Embedding the $SO(1,5)$ gravitational sector in $SO(1,17)$ and breaking the full gauge group yields Einstein gravity coupled to $SO(10)$. The Weyl-Majorana condition reduces the fermion family degeneracy to two, in contrast to the four families obtained in the $SO(2,16)$ conformal gravity unification.
Read the full abstract

Data Structures and Algorithms

Online Bin Packing with Per-Bin Maximum Delay

In online bin packing with delay, items of sizes in $(0,1]$ arrive over time and may wait before being packed into unit-capacity bins. A service selects a nonempty feasible batch of pending items and seals it in a new bin, at a service cost of one. We study the objective that adds one maximum waiting time for each sealed bin to the total service cost. Thus a bin's delay is determined by its oldest item, and younger items that fit can join it without increasing that delay. We give a randomized online algorithm with strict competitive ratio $2$ against an oblivious adversary, running in polynomial time in the ideal random-real model. It maintains one Next-Fit buffer and uses a single random silence threshold. Its analysis combines an offline-bin fragmentation bound with a two-unit saving per silence cluster. The ratio is tight for this policy; $b$ fair bits suffice for ratio $2+2^{-b}$. A fixed threshold gives deterministic ratio $1+\sqrt3$. The lower bounds $2$ and $e/(e-1)$ are attained when capacity is nonbinding. Offline, the problem minimizes the number of bins plus their total arrival span. We prove strong NP-hardness, an absolute approximation barrier of $3/2$, a deterministic polynomial $2$-approximation, and exact algorithms for several special cases. We also give stronger online ratios with exact endpoint-packing oracles and an AFPTAS for the fixed-length weighted endpoint subproblem. Finally, for Poisson arrivals of half-size items, we determine the offline and optimal causal long-run cost rates. Two natural asymptotic ratio measures coincide and have maximum $4/3$.
Read the full abstract

Statistical Methodology

Confidence regions for SAFE AI compliance scores

Artificial Intelligence trustworthiness scores are moving from research dashboards into compliance claims and procurement decisions. A compliance claim is a statistical decision based on the comparison between an estimated compliance metric and a set threshold. The recently proposed integrated SAFE AI metrics are based on three basic rank-graduation measures of accuracy, explainability and robustness, expressed on a common footing. The metrics are integrated into a single compliance score, so far reported as a point value. In this paper we propose to add an uncertainty layer to the SAFE AI compliance score. All component metrics are estimated on a shared test sample, so their errors covary. We estimate the full covariance of the component vector with a paired bootstrap, and quantify departures from the source-wise independence baseline suggested by the integrated-metrics framework's variability decomposition. A closed-form identity splits such departure into an uncertainty-weighted effective dimension and an exposure-weighted data-driven correlation, with both depending on the aggregator and the estimated uncertainty of the component metrics. The application of our proposal to both real and simulated data shows that the assumption of independence can either narrow or widen confidence bands, while the direction and magnitude of the effect depend on the aggregator type, machine learning model, and perturbation family. The resulting confidence region supports compliance decisions, which can be documented in a formal uncertainty certificate.
Read the full abstract

Logic in Computer Science

Step Recursion: Exact Depth Does Not Determine Algebraic Expressiveness

Step recursion is a form of bounded recursion in which each recursive call moves from an input y to a prescribed predecessor $\rho_g(y)$. Its depth $D_g(y)$ is the exact number of such moves needed to reach zero. A natural question is whether knowing this depth for every input determines the expressive power of the resulting function algebra. We prove that it does not. We first construct two simple generators with exactly the same depth map but different step-recursion algebras. One gives ordinary binary halving, $b(x)=2x+1$; the other is $p(0)=1$, $p(x)=x+2^{\lambda(x)}$ for $x>0$, where $\lambda$ is binary length. Although $D_p=D_b$ pointwise, the predecessor $\rho_p$ cannot be defined from any fixed-stride binary-halving descent at basis zero. Thus two recursion schemes may take exactly the same number of steps on every input and still have different expressive power. The phenomenon is much larger than this example. Whenever infinitely many depth levels allow more than one predecessor arrangement, a single exact depth profile supports $2^{\aleph_0}$ distinct step-recursion algebras over every countable basis containing zero and the projections. In the computable setting the corresponding effective family has exactly $\aleph_0$ distinct algebras. Hence exact recursion depth is an informative resource measure, but it is not a complete invariant: the geometry of predecessor choices inside each depth level carries additional algebraic information.
Read the full abstract

Computational Complexity

Step Recursion: Mixed Stride Spectra, Path Factorization, and Synchronization Geometry

Step recursion allows recursive computation to move through a canonical hierarchy in jumps, or strides. Suppose a function algebra is allowed to use several primitive stride lengths $L$. Composition immediately produces sums of these lengths, but it is not clear whether arbitrary nesting of mixed recursions can create any genuinely new canonical stride. We prove that it cannot: at every fixed canonical row $n\ge2$ and lower basis $m<n$, the canonical strides definable from $L$ are exactly the additive monoid $\langle L\rangle$ generated by $L$. The proof gives more information than membership alone. Every sufficiently high dependency path computing a canonical descent of stride $p$ carries total label weight exactly $p$, and the possible path signatures are precisely the additive factorizations of $p$ by the primitive stride labels. Thus the computation remembers both which strides are definable and how each one can be assembled. After saturation, inclusion between mixed-stride classes is exactly inclusion between their additive stride monoids, giving a concrete lattice description and finite certificates for inclusion. We then study several synchronized recursion clocks. Joint descent maps can encode arbitrary additive submonoids of $\mathbb N^r$, already giving continuum order complexity for $r=2$. Ordinary scalar observation, however, forgets correlations between the clocks and retains only the independent coordinate strides. This identifies precisely where synchronization information is preserved and where it collapses.
Read the full abstract

Machine Learning

A Computational Framework for Modelling Organisation-Level Semantic Identity from Longitudinal Textual Data

Can longitudinal texts produced by organisationally affiliated contributors reveal collective semantic differentiation and change without obscuring differences among those contributors? We develop a computational social science framework that combines established semantic and graph measures into interpretable organisational fingerprints and temporal profiles. We use organisation-level semantic identity to describe a corpus-dependent interpretation of recurring and changing features of affiliated textual output, rather than corporate culture or managerial intention. Drawing on an inherited semantic landscape, we examine 1,289 songs by 126 artists affiliated with four K-pop entertainment companies between 1997 and 2023. The evidence for differentiation is selective. After correction, all three novelty-related measures show artist-level differences, whereas the diversity and connectivity measures tested do not; only two company–community enrichments remain significant. Temporal trajectories reveal further distinctions between portfolios. Artist-grouped balanced accuracy reaches 0.334 against a chance baseline of 0.25, providing modest supporting evidence of affiliation-related signal rather than independent validation of identity. Robustness analyses support the presence of continuous semantic structure more strongly than precise rankings. Together, these findings show how collective regularities can coexist with shared content, contributor heterogeneity and change over time, while leaving cross-domain generality and broader organisational construct validity unresolved.
Read the full abstract

Complex Variables

Rigidity for local holomorphic isometric maps from the complex unit ball to reducible bounded symmetric domains

Let $U$ be a connected open subset of the complex unit ball $B^n$, $n\geq2$, and let $F_i: U \to \Omega_i$ be non-constant holomorphic maps into irreducible bounded symmetric domains. We prove that if $F=(F_1, \cdots, F_m)$ is a holomorphic isometric map from $U$ to the product of $\Omega_i$, then every $F_i$ extends to a proper holomorphic isometric embedding up to a positive constant from $B^n$ into $\Omega_i$.
Read the full abstract

Rings and Algebras

Failure of Ado-type theorems for preLie and postLie algebras

Over an arbitrary field of characteristic zero, we construct a two-dimensional preLie algebra and a two-dimensional postLie algebra that do not embed into the corresponding induced structures of any finite-dimensional preassociative or postassociative algebra, respectively.

Machine Learning

When Identical Rows Disagree: From Benchmark Identifiability to Replication-Robust Anomaly Detection

A released table is often treated as an i.i.d. sample, although its repeated rows may encode business frequency, repeated entities, joins, resampling, or extraction errors. We show that this ambiguity creates a hidden measurement layer with three consequences: feature-identical rows impose an attained evaluation ceiling, row-weighted AUROC is sensitive to replication, and row-trained detectors learn a multiplicity-size-biased law. An exact-row audit of all 690 OddBench datasets finds train-test overlap in 355, feature-identical label conflict in 147, and a test anomaly identical to a training normal in 137. Switching from row to support weighting changes AUROC by at least 0.05 on 50-61 datasets across four classical detector geometries. We introduce SCOUT (Support-Count Orthogonalized Unsupervised Testing), a factorized anomaly detector that separates replication-invariant support evidence from exposure-aware count evidence. Factorwise split-conformal calibration yields marginal false-positive-rate control, while the support channel is exactly invariant to arbitrary positive row replication. On 686 OddBench datasets and five seeds, support-only SCOUT is non-inferior to row-wise Isolation Forest in raw AUROC and improves replication-invariant AUROC. External normal-support evaluations track nominal false-positive levels, and four backbones remain exactly unchanged under controlled replication. Semi-synthetic interventions show that conditional count modeling helps materially only under strong rate heterogeneity. These results specify when multiplicity should be treated as signal, nuisance, or uninterpretable without additional information.
Read the full abstract

Algebraic Geometry

Stringy Irregular Hodge Numbers for Non-Simplicial Toric Landau–Ginzburg Data

We extend orbifold irregular Hodge theory from simplicial toric Landau–Ginzburg models to data defined by possibly non-simplicial stacky fans. For a Clarke mirror pair, we construct a filtered Koszul–de Rham complex over toric face rings. For simplicial ray-supported data, a strict filtered comparison identifies this complex with the Harder–Lee cellular residue complex, and hence it computes the orbifold irregular Hodge numbers. Using this complex, for a fixed possibly non-simplicial datum, we prove that the models obtained from its projective crepant simplicial subdivisions have the same orbifold irregular Hodge polynomial whenever the potential is nondegenerate at infinity. This common value defines the stringy irregular Hodge numbers of the original datum.
Read the full abstract

Artificial Intelligence

PartHackBench: Certified Equal-Progress Stress Tests for Partial-Credit Tool-Agent Evaluation

Long-horizon tool agents often make useful progress without reaching terminal success, motivating partial-credit evaluation. Yet evaluators may reward milestones that were temporary, later reversed, or not attributable to the evaluated agent. Comparing an honest trajectory with a higher-scoring adversarial one is inconclusive if the latter made more genuine progress. We introduce PartHackBench, a controlled methodology that removes this confound. A private certifier admits a pair only when its trajectories match component-wise in both current-state predicate satisfaction and standardized agent attribution; score inflation, defined as f(A) - f(H), is measured only afterward. In 18 sealed held-out tasks in PB-CSTE, the frozen historical-target run produced matched adversaries for 15 tasks. Historical credit yielded mean inflation of .252, conditional attack success of 10/15, end-to-end yield of 10/18, and detected none of 14 strict rollbacks. Semantic LLM judges were more resistant but remained vulnerable, especially under evaluator-targeted attacks, while PB-CSTE current-state controls, defined as exact functions of the certified components, yielded zero inflation by construction. PartHackBench thus provides a certified control for testing whether evaluator credit changes while all benchmark-defined task-relevant progress remains fixed.
Read the full abstract

Dynamical Systems

A Chern-Class Obstruction to Measurable Eigensections of Analytic Quasi-Periodic Bundle Cocycles

We study linear cocycles over ergodic translations of the two-torus acting on Hermitian vector bundles described by unitary sewing matrices. For a continuous invariant complex line subbundle $L$, we prove that $c_1(L) \ne 0$ obstructs every nonzero measurable eigensection under a measurable phase-coboundary condition. The abstract obstruction holds in every rank, while for rank-two analytic cocycles with a dominated splitting over a Diophantine translation the phase condition is automatic. In that regime the criterion is exact: $L$ carries a nonzero measurable eigensection if and only if $c_1(L)=0$ and the winding vector of the multiplier phase vanishes. When these conditions hold, the realized eigenvalues form a dense coset on a circle; for each such eigenvalue, the eigenspace carried by $L$ is one-dimensional and has a real-analytic nowhere-vanishing generator. We also construct a Liouville example showing that the Diophantine hypothesis cannot be removed. The proof converts the Chern number into a magnetic charge on the universal cover and applies covariance for irrational magnetic translations.
Read the full abstract

Rings and Algebras

Differential perm bialgebras, differential perm Yang-Baxter equation, and their applications in related bialgebra structures

We introduce the differential perm bialgebras by generalizing the study of perm bialgebras to the context of differential perm algebras. They are characterized by the Manin triple of differential perm algebras as well as the matched pairs of differential perm algebras. The quasi-triangular (resp. triangular, factorizable) differential perm bialgebra are introduced. In particular, symmetric solutions of an analogue of perm Yang-Baxter equation in differential perm algebras provide triangular differential perm bialgebras, whereas in turn the notions of $\mathcal{O}$-operators of differential perm algebras. As an application, the relationship between differential perm bialgebras and related bialgebra structures is discussed in detail. We show that there is a natural differential infinitesimal bialgebra structure on the tensor product of a differential perm bialgebra and a quadratic Zinbiel algebra, and this differential infinitesimal bialgebra structure is quasi-triangular (resp. triangular, factorizable) if the original differential perm bialgebra is quasi-triangular (resp. triangular, factorizable). Starting from a differential perm bialgebras with certain conditions, we can construct Novikov bialgebras, and these Novikov bialgebras can also be implemented by using the tensor product of diNovikov bialgebras and quadratic Zinbiel algebras.
Read the full abstract

Representation Theory

A Cotangent Model for the Next-to-Minimal Orbit of E7

Let $\mathcal O_{2A_1}$ be the next-to-minimal nilpotent orbit of the complex simple Lie algebra of type $E_7$. Let $\mathfrak p_7$ be the maximal parabolic subalgebra corresponding to the simple root $\alpha_7$, and let $\mathfrak u^+$ be its nilpotent radical. We prove that the affinization $T^*(\mathcal O_{2A_1}\cap\mathfrak u^+)^{\mathrm{aff}}$ of the cotangent bundle of $\mathcal O_{2A_1}\cap\mathfrak u^+$ is isomorphic to the nilpotent orbit closure $\overline{\mathcal O}_{2A_1}$ as affine varieties.
Read the full abstract

Combinatorics

Every graph is eventually Turán-good and Turán-stable

Let $T_r(n)$ denote the complete $r$-partite graph on $n$ vertices whose part sizes differ by at most one. A graph $H$ is called $K_{r+1}$-Turán-good if, for all sufficiently large $n$, the graph $T_r(n)$ contains the maximum number of copies of $H$ among all $n$-vertex $K_{r+1}$-free graphs. We say that $H$ is $K_{r+1}$-Turán-stable if, for every $\varepsilon>0$, there exist $\delta>0$ and $n_0$ such that, whenever $n\ge n_0$ and an $n$-vertex $K_{r+1}$-free graph $G$ contains at least $\mathrm{ex}(n,H,K_{r+1})-\delta n^{v(H)}$ copies of $H$, the edit distance between $G$ and $T_r(n)$ is at most $\varepsilon n^2$, where $v(H)$ denotes the order of $H$. In this paper, we prove that every graph $H$ is both $K_{r+1}$-Turán-good and $K_{r+1}$-Turán-stable for $r\ge 4v(H)^3+11v(H)^2$. This strengthens results of Morrison, Nir, Norin, Rzążewski, and Wesolek~[J. Combin. Theory Ser. B, 2023] and Gerbner and Hama Karim~[J. Graph Theory, 2024], and gives a positive answer to a question of Morrison, Nir, Norin, Rzążewski, and Wesolek. We also prove that $\inj(H,G)\le \inj(H,T_r(n))$ for every $n$-vertex $K_{r+1}$-free graph $G$ and for $r\ge 40v(H)^3$, where $\inj(H,G)$ denotes the number of injective homomorphisms from $H$ to $G$. Finally, we give a negative answer to another question of Morrison, Nir, Norin, Rzążewski, and Wesolek.
Read the full abstract

Information Retrieval

The Recall Ceiling of LLM Recommendation Reranking

Some LLM-based recommendation rerankers are evaluated under an oracle protocol that guarantees the ground-truth item is present in the scored set, either by injecting it into the candidate list or by scoring it against sampled negatives. Across three primary Amazon datasets, we show that this protocol overestimates realistic NDCG@10 by 92–95%. The cause is a recall ceiling: realistic retrieval covers only 2–19% of relevant items at $K=100$ across eight datasets in three domains, imposing a deterministic upper bound on any closed-candidate reranker's top-$k$ NDCG. Under leave-one-out evaluation, $\mathbb{E}[\mathrm{NDCG}@k] \leq \mathrm{Recall}@|W_\pi|$, where $W_\pi$ is the reranker's candidate window. Under realistic retrieval, none of the tested optimisation strategies significantly improves over the collaborative-filtering baseline on our primary Amazon datasets. These strategies include prompt engineering, model scaling over a 168$\times$ parameter range, sequential models, supervised neural rerankers, LoRA fine-tuning, hybrid retrieval, score-aware prompting, and LLM+CF fusion. Text-aware retrieval increases recall on one dataset but does not improve end-to-end NDCG, while providing upstream CF scores mainly makes the LLM reproduce the CF order. We therefore propose the Recall-Aware Evaluation Protocol (RAEP): first classify the retrieval-recall regime, then evaluate reranking where the ceiling permits meaningful differentiation. In the low-recall regimes measured here, improving retrieval is more consequential than increasing reranker sophistication; this ordering need not hold in production systems with higher recall, richer features, or online feedback.
Read the full abstract