2026-09论文中文摘要
RAISE：通过符号评估强化LLMs中的访问控制策略合成
RAISE: Reinforcing Access Control Policy Synthesis in LLMs via Symbolic Evaluation
arXiv 2609.33796 · 2026-09-29氢-柴油双燃料发动机实时运行的模型预测控制的深度行为克隆
Deep Behaviour Cloning of Model Predictive Control for Real-Time Operation of a Hydrogen-Diesel Dual-Fuel Engine
arXiv 2609.33795 · 2026-09-29高阶超丰数
Higher-order colossally abundant numbers
arXiv 2609.33794 · 2026-09-29两性纳米颗粒表面的电荷调控介导相互作用
Charge Regulation Mediated Interaction of Amphoteric Nanoparticle Surfaces
arXiv 2609.33793 · 2026-09-29椭圆曲线在 $\mathbb{F}_q(t)$ 上的秩在剩余类中的稳定性
Rank stability of elliptic curves over $\mathbb{F}_q(t)$ in residue classes
arXiv 2609.33792 · 2026-09-29我们真的需要KL散度用于大型语言模型的同策略蒸馏吗？
Do We Really Need KL Divergence for On-Policy Distillation of Large Language Models?
arXiv 2609.33791 · 2026-09-29多孔径PPG与MAPIS：空间光学与时间相干场
Multi-Aperture PPG with MAPIS: Spatial Optical and Temporal Coherence Fields
arXiv 2609.33790 · 2026-09-29指数次椭圆调和映射的不稳定性与Morse指标
Instability and Morse Index of Exponentially Subelliptic Harmonic Maps
arXiv 2609.33789 · 2026-09-29横截面 Siegel-Veech 变换的矩
Moments of the Cross-Sectional Siegel-Veech Transform
arXiv 2609.33788 · 2026-09-29LA-CPD：面向人机合著篇章归属分割的局部证据感知变点检测
LA-CPD: Local-Evidence-Aware Change-Point Detection for Human-LLM Authorship Segmentation
arXiv 2609.33787 · 2026-09-29你的不确定性图错了，还是它的目标错了？Tweedie对角线的精确诊断，以及一种无梯度的替代方案
Is your uncertainty map wrong, or is its target? Exact diagnostics for the Tweedie diagonal, and a gradient-free alternative
arXiv 2609.33786 · 2026-09-29$S^6$ 上的一个双参数复结构族
A two-parameter family of complex structures on $S^6$
arXiv 2609.33785 · 2026-09-29关于机理上可访问的共聚物序列
On mechanistically accessible copolymer sequences
arXiv 2609.33784 · 2026-09-29有限弦长多孔翼型可压缩非定常空气动力学
Compressible unsteady aerodynamics of finite-chord porous aerofoils
arXiv 2609.33783 · 2026-09-29通过联络与留数的曲面等变黎曼-罗赫定理
Equivariant Riemann-Roch for Surfaces via Connections and Residues
arXiv 2609.33782 · 2026-09-29惊人成功，反复失败：熵引导的信用分配用于LLM推理中的探索
Surprising Success, Repeated Failure: Entropy-Guided Credit Assignment for Exploration in LLM Reasoning
arXiv 2609.33781 · 2026-09-29选择多样化的SFT轨迹提升后强化学习泛化能力
Selecting Diverse SFT Traces Improves Post-RL Generalization
arXiv 2609.33780 · 2026-09-29证据-推理重建：当证据被回忆但推理出错时
Evidence-Inference Reconstruction: When The Evidence Is Recalled But The Reasoning Goes Wrong
arXiv 2609.33778 · 2026-09-29低能量子子空间学习的标度理论
Scaling Theory for Learning Low-Energy Quantum Subspaces
arXiv 2609.33777 · 2026-09-29含钛氧化物中第一性原理钛K边X射线吸收光谱模拟的基准测试
Benchmark of First-Principles Titanium K-Edge X-Ray Absorption Spectral Simulations on Titanium-containing Oxides
arXiv 2609.33776 · 2026-09-29短程质量次临界 Choquard 方程的散射
Scattering for Short-Range Mass-Subcritical Choquard Equations
arXiv 2609.33775 · 2026-09-29令牌变化，结构持久：生成语音的频谱水印
Tokens Change, Structure Endures: Spectral Watermarking for Generated Speech
arXiv 2609.33774 · 2026-09-29学习破解评判者的策略
Learning Strategies to Break Judges
arXiv 2609.33773 · 2026-09-29二块奇数分拆函数
The two-block odd partition function
arXiv 2609.33770 · 2026-09-29M3-Score：将保真度、记忆和覆盖度作为评估生成式放射学图像模型的独立轴
M3-Score: Fidelity, Memorization and Coverage as Separate Axes for Evaluating Generative Radiology Image Models
arXiv 2609.33769 · 2026-09-29DEALS：面向多智能体LLM系统的去中心化专业知识感知负载服务
DEALS: Decentralized Expertise-Aware Load Serving for Multi-Agent LLM Systems
arXiv 2609.33768 · 2026-09-29驯服希腊字母：具有归纳偏见的期权组合
Taming the Greeks: Option Portfolios with Inductive Biases
arXiv 2609.33767 · 2026-09-29互反多项式为单项式的充要条件
Necessary and sufficient condition for reciprocal polynomials to be monogenic
arXiv 2609.33766 · 2026-09-29主转向子空间：冻结生成式机器人策略的在线自适应
Principal Steering Subspaces for Online Adaptation of Frozen Generative Robot Policies
arXiv 2609.33765 · 2026-09-29超越固定特征：异质性下节点表示对架构的依赖性敏感性
Beyond Fixed Features: Architecture-Dependent Sensitivity to Node Representations under Heterophily
arXiv 2609.33764 · 2026-09-29SecProbe：编码智能体在网络安全漏洞上的自适应评估
SecProbe: Adaptive Evaluation of Coding Agents on Cybersecurity Vulnerabilities
arXiv 2609.33763 · 2026-09-29EfficientAgent：是什么让KV缓存卸载对并发智能体有效？
EfficientAgent: What Makes KV Cache Offloading Work for Concurrent Agents?
arXiv 2609.33762 · 2026-09-29StatD2GAN：当校准在合成天气序列的留出评估中掩盖生成器质量
StatD2GAN: When Calibration Masks Generator Quality in Held-Out Evaluation of Synthetic Weather Sequences
arXiv 2609.33761 · 2026-09-29位置并非事实：KV缓存与记忆之间的错配
Positions Are Not Facts: The Mismatch Between KV Caches and Memory
arXiv 2609.33759 · 2026-09-29ENet-GP：统一文档图像恢复
ENet-GP: Unified Document Image Restoration
arXiv 2609.33758 · 2026-09-29YuE2：在前沿质量上统一符号音乐与音频音乐生成
YuE2: Unifying Symbolic and Audio Music Generation at Frontier Quality
arXiv 2609.33757 · 2026-09-29类灯群的分析、几何与度量方面
Analytic, geometric and measured aspects of lamplighter-like groups
arXiv 2609.33756 · 2026-09-29基于Transformer的神经波束成形，用于智能低功耗可听戴设备的实时语音增强
Transformer-based Neural Beamforming for Real-Time Speech Enhancement on Smart Low-Power Hearable Devices
arXiv 2609.33755 · 2026-09-29并发编码信号复用测距用于半双工异步网络
Concurrent Coded Signal-Multiplexing Ranging for Half-Duplex Asynchronous Networks
arXiv 2609.33753 · 2026-09-29随机谐波约束下布朗运动的非平稳统计与能量学
Non-stationary Statistics and Energetics of Brownian Motion under Stochastic Harmonic Confinement
arXiv 2609.33751 · 2026-09-29计算专业学生如何与大型语言模型交互用于学习目的的实证研究
An Empirical Study on how Computing Students Interact with Large Language Models for Learning Purposes
arXiv 2609.33750 · 2026-09-29ResDiffFRG：用于多种适当面部反应生成的残差扩散
ResDiffFRG: Residual Diffusion for Multiple Appropriate Facial Reaction Generation
arXiv 2609.33749 · 2026-09-29任务感知的可微逻辑门网络离散化
Task-Aware Discretization of Differentiable Logic Gate Networks
arXiv 2609.33747 · 2026-09-29PQ-HSA：复用乘积量化分数实现混合稀疏-近似注意力
PQ-HSA: Reusing Product-Quantized Scores for Hybrid Sparse-Approximate Attention
arXiv 2609.33746 · 2026-09-29超越波束：生成式推荐中的构造性修复与候选补全
Beyond the Beam: Constructive Repair and Candidate Completion for Generative Recommendation
arXiv 2609.33745 · 2026-09-29斐波那契数不是3-可访问的
The Fibonacci numbers are not 3-accessible
arXiv 2609.33744 · 2026-09-29大型语言模型拒绝行为研究中的随机性考量
Accounting for Stochasticity in Studies of Large Language Model Refusal
arXiv 2609.33743 · 2026-09-29DuraS2ST：面向时长对齐的语音到语音翻译的思维链与强化学习
DuraS2ST: Chain-of-Thought and Reinforcement Learning for Duration-Aligned Speech-to-Speech Translation
arXiv 2609.33742 · 2026-09-29多边际逆最优传输用于对比学习：基于显式锚点-正样本-负样本耦合
Multi-Marginal Inverse Optimal Transport for Contrastive Learning Via Explicit Anchor-Positive-Negative Coupling
arXiv 2609.33741 · 2026-09-29加权样条扩展网络与分布平衡用于连续处理效应估计
Weighted Spline-Expanded Networks with Distributional Balancing for Continuous Treatment Effects
arXiv 2609.33740 · 2026-09-29
