arXiv subjects

Lishan Yu

Publications and source records attributed to Lishan Yu.

5 recordsLinked to original sources

arXiv2026

OpenBrain: An Auditable Generated-Label Release for Whole-Brain MRI Parcellation

Automatic tools can now generate whole-brain MRI parcellations at cohort scale, but generated-label releases often provide little evidence about which cases are reliable enough to reuse. We introduce OpenBrain, a confidence-aware release of 35,838 provenance-verified OpenNeuro T1w cases from 607 source datasets. OpenBrain is organized as an auditable case-record release rather than a directory of standalone segmentation files: each record links a primary parcellation, two committee parcellations, source/license provenance, raw QC-7 measurements, percentile risks, a combined risk score R, and a confidence tier. QC-7 is a reference-free reliability panel over label geometry, image-label consistency, and committee disagreement, so users can filter, weight, audit, or re-threshold generated labels without rerunning inference. We evaluate this released evidence under a frozen protocol. Across external cohorts, lower QC-7 risk corresponds to higher generated-label quality; in fixed-budget training, lower-risk selections improve paired macro Dice over random selection by +3.2 to +3.7 percentage points. OpenBrain also exposes raw axes, committee outputs, provenance, and datasheet documentation, so confidence claims remain inspectable after download. It does not turn generated labels into manual annotations or clinical-grade segmentations. Its contribution is a large open generated-parcellation substrate, a reusable case-level risk-evidence layer, and an external validation protocol for release-time QC. Dataset: https://huggingface.co/datasets/lanqz7766/OpenBrain. Code: https://github.com/lanqz7766/openbrain.

cs.CV↗

arXiv2026

Pellarin Values from Cauchy Coordinates on Arbitrary Curves

Let $A$ be the ring of functions regular outside a closed point of degree $\mathrm{N}$ on a smooth projective curve over $\mathbb{F}_q$. We express the normalized weight-one ideal Pellarin value of a rank-one Drinfeld $A$-module as an ordinary Hilbert-class trace of rational Cauchy differentials. A pairing on projective period lattices identifies the differential associated with each ideal class and computes its coefficients by finite torsion traces. For a nonfree lattice, the formula uses an isogeny with arbitrary nonzero derivative to a module with free period lattice. Each ideal-class subseries extends regularly to the characteristic point with value one, so the full series takes the value $|\operatorname{Cl}(A)|$. For every rank-one module, the quotient of the Anderson generating differential by the Drinfeld exponential is a meromorphic lattice sum whose value at zero is the Cauchy differential. Its Taylor coefficients give the higher-weight lattice differentials. For sign-normalized Hayes modules with free period lattice, an ordinary Hilbert-class trace after projection to powers divisible by $q^{\mathrm{N}}-1$ generates ideal values at $n\equiv1\pmod{q^{\mathrm{N}}-1}$. An algebraic parameter rescaling gives a trace formula for character-twisted values at $n\equiv1\pmod{q-1}$. The exact coefficient fields explain the difference between these weight ranges. For sign-normalized modules, decomposition by ideal classes also relates the element-type and ideal-type series at every positive weight. Explicit computations include the Hu--Huang--Yau module with nonprincipal period lattice and an elliptic curve with an infinite point of degree three, including the holomorphic term of its Cauchy differential. At rational infinity, the elliptic formula recovers the Green--Papanikolas identity.

math.NT↗

arXiv2026

Pellarin's Identity, Carlitz Period, and Anderson Generating Functions over Arbitrary Curves

Let $A$ be the coefficient ring of a smooth projective curve over $\mathbb{F}_q$ with a closed point at infinity of arbitrary degree $\mathrm{N}=°(\infty)\ge 1$, and let $φ$ be a rank-one Drinfeld $A$-module. In this paper, we prove three main results concerning the module of special functions and its relation to Pellarin's $\mathcal L(1)$-series. First, the module $\mathrm{sf}(φ)$ of special functions contains a Tate-algebra unit exactly when it is free of rank one, equivalently when its period lattice is isomorphic to the module of regular differentials. This settles the Gazda--Maurischat conjecture. The proof evaluates Cauchy kernels; on the period lattice all such evaluations agree, and their common characteristic residue recovers the period. Second, for suitable isogeny data to a principal-period target $ψ$, a strictly normalized shtuka product, together with one characteristic residue, gives an explicit formula for the fundamental period $\tildeπ$ (the generator of the free lattice $Λ_ψ$). The same defect equation also yields the residue formula for the Drinfeld logarithm. Third, the first Frobenius twist of the normalized shtuka differential is identified with a pairing built from finite torsion traces, whose coefficients are finite étale traces over the field of definition. As an application, Pellarin's $\mathcal L(1)$-series is expressed exactly as the first Frobenius twist of the generating differential at a fundamental period. A notable feature of these results is that they hold for arbitrary $\mathrm{N} \geq 1$, and the source period lattice need not be free.

math.NT↗

arXiv2026

EvoTreeNAD: Genealogy-Guided Evolution for LLM-Driven Neural Architecture Discovery

AI-driven scientific discovery accelerates research by autonomously developing solutions and designs. Large language model (LLM) agents support this process through iterative generation and evaluation. Yet these iterations alone do not ensure cumulative progress or establish which directions to pursue next. Costly evaluation further constrains the scope of exploration. Neural architecture discovery brings these challenges together, coupling open-ended design with resource-intensive experimentation. We introduce EvoTreeNAD, a genealogy-guided evolutionary algorithm that constructs trainable architectures without a supplied seed or a hand-specified search space. Starting from an empty root, it grows a persistent genealogy in which each new node represents a complete architecture. Top-percentile values computed from each node and its descendants guide lineage selection. Using the selected design history, an Idea Agent proposes a variant and a Code Agent implements it. Each evaluated variant becomes a child node, expanding the genealogy while providing evidence for subsequent lineage selection. Our theoretical analysis establishes the existence of stationary variation regimes as the genealogy grows. Under specified variation assumptions, sustained top-percentile family values quantify the probability of generating high-reward architectures in these regimes. EvoTreeNAD discovers architectures that outperform the compared NAS and NAD baselines, achieving CIFAR-10/100 test errors of $2.05{\pm}0.06\%$ and $15.09{\pm}0.22\%$. On all six MedMNIST-v2 tasks, the discovered architectures surpass the strongest listed baselines. A controlled CIFAR-10 study further shows that EvoTreeNAD outperforms direct generation, best-of-$N$ greedy continuation, and full-family-mean routing.

cs.NE↗

arXiv2026

Geometric Realization of Finite Residue Casimirs and Weil Operators via Szegő Kernels

For a smooth projective curve over a finite field with a fixed point at infinity, we establish a universal correspondence between finite residue duality and geometric kernel functions. We prove that the finite residue Casimir tensor is realized geometrically as the intrinsic principal part of the normalized Szegő kernel for any acyclic line bundle, and equivalently that this kernel acts as a reproducing kernel for the finite residue pairing, in exact analogy with the classical Cauchy integral formula. In the polynomial case these equivalent descriptions yield a closed formula involving the rank-two Weil operator, identified with the classical divided difference, recovering a remainder identity of Hu--Ou. These results provide a geometric foundation for the study of Anderson generating functions and the Weil pairing for Drinfeld modules.

math.NT↗