2026-07论文中文摘要
两栖动物形态色素沉着的磁敏感性依赖于光和眼睛，且与自由基对机制一致
Magnetosensitivity of amphibian morphological pigmentation is light- and eye-dependent and consistent with the radical pair mechanism
arXiv 2607.20513 · 2026-07-24μ子学习理解的有效成分
The Active Ingredient in Muon's Grokking
arXiv 2607.20512 · 2026-07-24SiGMA：用于多模态持续指令微调的符号引导合并与适配
SiGMA: Sign-Guided Merging and Adaptation for Multimodal Continual Instruction Tuning
arXiv 2607.20511 · 2026-07-24兰伯特W扭结神经冲动的超对称配对
Supersymmetric pairing of Lambert W-kink nerve impulses
arXiv 2607.20509 · 2026-07-24生物物理学中的哈密顿动力学与基本现象：综述
Hamiltonian Dynamics and Fundamental Phenomena in Biophysics: A Review
arXiv 2607.20508 · 2026-07-24优化基于超图的RAG：迈向更好的事实提取和块检索
Optimizing Hypergraph-Based RAG: Toward Better Fact Extraction and Chunk Retrieval
arXiv 2607.20506 · 2026-07-24HERMES：用于信号交叉口交通冲突预测的异构边缘关系多头嵌入式SSM注意力模型
HERMES: Heterogeneous Edge-Relational Multi-Head Embedded SSM Attention for Traffic Conflict Prediction at Signalized Intersections
arXiv 2607.20505 · 2026-07-24反应坐标相关的非马尔可夫摩擦支配蛋白质折叠动力学
Reaction-Coordinate-Dependent Non-Markovian Friction Governs Protein-Folding Dynamics
arXiv 2607.20504 · 2026-07-24LeanFlow：工作流驱动的精益自动形式化案例研究
LeanFlow: A Case Study in Workflow-Driven Lean Autoformalization
arXiv 2607.20503 · 2026-07-24诱导因式分解概率分布的可比性
Inducing Comparability of Factorised Probability Distributions
arXiv 2607.20502 · 2026-07-24MKEvolve：用于内核代码生成的模块化多智能体框架
MKEvolve: A Modular Multi-Agent Framework for Kernel Code Generation
arXiv 2607.20501 · 2026-07-24FlowEdit：针对涉及冲突的不适定问题的大语言模型推理流的信息论控制
FlowEdit: Information-Theoretic Control of LLM Reasoning Flows for Ill-posed Problems Involving Conflicts
arXiv 2607.20500 · 2026-07-24ExecuGraph：一种用于通过大语言模型进行可靠后端代码合成的多智能体、基于执行的框架
ExecuGraph: A Multi-Agent, Execution-Grounded Framework for Reliable Backend Code Synthesis with Large Language Models
arXiv 2607.20499 · 2026-07-24AISE-Bench：用于学术知识图谱信息检索的全周期精选基准测试
AISE-Bench: A Full-Cycle Curated Benchmark for Information Seeking on Academic Knowledge Graphs
arXiv 2607.20498 · 2026-07-24动作电位与孤子
Action potentials and solitons
arXiv 2607.20496 · 2026-07-24多智能体系统的工作负载感知缓存
Workload-Aware Caching for Multi-Agent Systems
arXiv 2607.20495 · 2026-07-24从正则表达式中分离大语言模型对齐：对抗性变异下的零覆盖率和度量相关散度
Isolating LLM Alignment from Regex: Zero Coverage and Metric-Dependent Divergence Under Adversarial Mutation
arXiv 2607.20494 · 2026-07-24基于注意力的经验回放框架用于不可知时间序列预测模型的持续学习
Attention-based Experience Replay Framework for Continual Learning of Agnostic Time Series Forecasting Models
arXiv 2607.20493 · 2026-07-24CRAWO：用于自适应工作负载编排的定制资源
CRAWO: Custom Resources for Adaptive Workload Orchestration
arXiv 2607.20490 · 2026-07-24自主拓扑变异：具有能力、状态和影子不变量的多智能体大语言模型系统的安全运行时重组
Autonomous Topology Mutation: Safe Runtime Restructuring for Multi-Agent LLM Systems with Capability, State, and Shadow Invariants
arXiv 2607.20488 · 2026-07-24OPTScientist：用于Transformer预训练的类型化优化器程序的多智能体发现
OPTScientist: Multi-Agent Discovery of Typed Optimizer Programs for Transformer Pretraining
arXiv 2607.20486 · 2026-07-24语言模型与现实世界用户期望的期望对齐
Expectation Alignment of Language Models for Real-World User Expectations
arXiv 2607.20485 · 2026-07-24问题出在频谱中：通过拓扑正则化侧路径减轻大语言模型中的表示坍缩
The Devil is in the Spectrum: Mitigating Representation Collapse in LLMs via Topologically Regularized Side-Path
arXiv 2607.20484 · 2026-07-24自回归语言模型的可处理分层控制
Tractable Hierarchical Control of Autoregressive Language Models
arXiv 2607.20483 · 2026-07-24无需训练的路由：通过可靠性门控实现可控比例的大语言模型卸载
Routing Without Training: Controllable-Ratio LLM Offloading via Reliability Gating
arXiv 2607.20481 · 2026-07-24通过约束多源推理实现配电系统中可扩展的拓扑推理
Enabling Scalable Topology Inference in Distribution Systems via Constrained Multi-Source Inference
arXiv 2607.20480 · 2026-07-24超越说谎者基准：谎言类型、深度和稀疏性对大语言模型中欺骗检测的影响
Beyond Liars' Bench: The Impact of Lie Typology, Depth, and Sparsity on Deception Detection in LLMs
arXiv 2607.20479 · 2026-07-24用于基础设施即代码生成的智能语言模型的验证优先评估
Verifier-First Evaluation of Agentic LLMs for Infrastructure-as-Code Generation
arXiv 2607.20478 · 2026-07-24半监督文本属性图蒸馏
Semi-Supervised Text-Attributed Graph Distillation
arXiv 2607.20477 · 2026-07-24在多传感器物理危害评估中对大语言模型进行基准测试
Benchmarking Large Language Models on Multi-Sensor Physical Hazard Assessment
arXiv 2607.20476 · 2026-07-24SonicSampler：用于语言模型采样和推测性验证的统一瓦片感知内核
SonicSampler: Unified Tile-Aware Kernels for LLM Sampling and Speculative Verification
arXiv 2607.20475 · 2026-07-24大语言模型中的不完整提示越狱
Incomplete Prompt Jailbreaks in Large Language Models
arXiv 2607.20473 · 2026-07-24鲁棒评论家：防御大语言模型免受多轮攻击
Robust Critics: Defending LLMs Against Multi-Turn Attacks
arXiv 2607.20472 · 2026-07-24PlanE：基于抽取式大语言模型的数据、调优和推理的元规划
PlanE: Meta Planning of Data, Tuning, and Inference for Extractive-based LLMs
arXiv 2607.20470 · 2026-07-24DecodeShare：追踪大语言模型解码时刻决策的共享子空间
DecodeShare: Tracing the Shared Subspace of LLM Decode-Time Decisions
arXiv 2607.20469 · 2026-07-24InferenceBench：用于通过人工智能代理进行开放式大语言模型推理优化的基准测试
InferenceBench: A Benchmark for Open-Ended LLM Inference Optimization by AI Agents
arXiv 2607.20468 · 2026-07-24DC-Leap：通过草稿引导的连续跳跃解码实现无训练的dLLMs加速
DC-Leap: Training-Free Acceleration of dLLMs via Draft-Guided Contiguous Leaping Decoding
arXiv 2607.20467 · 2026-07-24JAXBench：自主TPU内核优化基准测试
JAXBench: Benchmarking Autonomous TPU Kernel Optimization
arXiv 2607.20466 · 2026-07-24数据准备基准：评估作为训练数据准备者的大语言模型
DataPrep-Bench: Benchmarking LLMs as Training Data Preparators
arXiv 2607.20465 · 2026-07-24随机采样在认知上是浅薄的：大语言模型中温度变化与模型多样性之间的维度差距
Stochastic Sampling is Epistemically Shallow: The Dimensionality Gap Between Temperature Variation and Model Diversity in LLMs
arXiv 2607.20464 · 2026-07-24ClickGuard：利用信息性度量和大语言模型检测和揭露标题党新闻
ClickGuard: Detecting and Spoiling Clickbait News with Informativeness Measures and Large Language Models
arXiv 2607.20463 · 2026-07-24标记错误症状：评估医学文本中的大语言模型水印
Marking the Wrong Symptoms: Evaluating LLM Watermarks in Medical Texts
arXiv 2607.20462 · 2026-07-24自然语言中的效价能反映道德吗？一项初步标注研究
Can Valence Reflect Morality in Natural Language? A Preliminary Annotation Study
arXiv 2607.20461 · 2026-07-24Instruct-FD：你的全双工语音系统能遵循轮流发言指令吗？
Instruct-FD: Can Your Full-Duplex Speech System Follow Turn-Taking Instructions?
arXiv 2607.20460 · 2026-07-24THOR：用于多跳问答的θ-γ分层振荡推理框架
THOR: A Theta-Gamma Hierarchical Oscillatory Reasoning Framework for Multi-hop QA
arXiv 2607.20459 · 2026-07-24CAMeR：用于大语言模型智能体中自适应记忆保留的关键词门控混合激活
CAMeR: Keyword-Gated Hybrid Activation for Adaptive Memory Retention in LLM Agents
arXiv 2607.20458 · 2026-07-24Learn2Zinc：微调小型语言模型以实现MiniZinc中的文本到模型翻译
Learn2Zinc: Fine-tuning Small Language Models for Text-to-Model Translation in MiniZinc
arXiv 2607.20456 · 2026-07-24RE-AD：数据标注的实时需求遵循
RE-AD: Real-Time Requirement Adherence for Data Labeling
arXiv 2607.20455 · 2026-07-24前沿大语言模型中的响应漂移
Response drift across frontier large language models
arXiv 2607.20454 · 2026-07-24一种用于将大语言模型零样本适应谵妄预测的知识注入框架
A Knowledge-Injection Framework for Zero-Shot Adaptation of LLMs to Delirium Prediction
arXiv 2607.20453 · 2026-07-24
