daily_AIOR_report

2026-05-20

WavFlow: Audio Generation in Waveform Space

• 英文标题: WavFlow: Audio Generation in Waveform Space
• 来源: 🤗 HuggingFace ⭐ 5/10
• 作者: Feiyan Zhou, Luyuan Wang, Shoufa Chen

• 论文链接: https://huggingface.co/papers/2605.18749

Symmetry-Compatible Principle for Optimizer Design: Embeddings, LM Heads, SwiGLU MLPs, and MoE Routers

• 英文标题: Symmetry-Compatible Principle for Optimizer Design: Embeddings, LM Heads, SwiGLU MLPs, and MoE Routers
• 来源: 🤗 HuggingFace ⭐ 5/10
• 作者: Tim Tsz-Kit Lau, Weijie Su

• 论文链接: https://huggingface.co/papers/2605.18106

DexHoldem: Playing Texas Hold’em with Dexterous Embodied System

• 英文标题: DexHoldem: Playing Texas Hold’em with Dexterous Embodied System
• 来源: 🤗 HuggingFace ⭐ 5/10
• 作者: Feng Chen, Tianzhe Chu, Li Sun

• 论文链接: https://huggingface.co/papers/2605.18727

TopoPrimer: The Missing Topological Context in Forecasting Models

• 英文标题: TopoPrimer: The Missing Topological Context in Forecasting Models
• 来源: 🤗 HuggingFace ⭐ 5/10
• 作者: Zara Zetlin, Kayhan Moharreri, Maria Safi

• 论文链接: https://huggingface.co/papers/2605.15035

SafeDiffusion-R1: Online Reward Steering for Safe Diffusion Post-Training

• 英文标题: SafeDiffusion-R1: Online Reward Steering for Safe Diffusion Post-Training
• 来源: 🤗 HuggingFace ⭐ 5/10
• 作者: Komal Kumar, Ankan Deria, Abhishek Basu

• 论文链接: https://huggingface.co/papers/2605.18719

AgentKernelArena: Generalization-Aware Benchmarking of GPU Kernel Optimization Agents

• 英文标题: AgentKernelArena: Generalization-Aware Benchmarking of GPU Kernel Optimization Agents
• 来源: 🤗 HuggingFace ⭐ 5/10
• 作者: Sharareh Younesian, Wenwen Ouyang, Sina Rafati

• 论文链接: https://huggingface.co/papers/2605.16819

AtlasVA: Self-Evolving Visual Skill Memory for Teacher-Free VLM Agents

• 英文标题: AtlasVA: Self-Evolving Visual Skill Memory for Teacher-Free VLM Agents
• 来源: 🤗 HuggingFace ⭐ 5/10
• 作者: Pan Wang, Yihao Hu, Xiujin Liu

• 论文链接: https://huggingface.co/papers/2605.17933

Code as Agent Harness

• 英文标题: Code as Agent Harness
• 来源: 🤗 HuggingFace ⭐ 5/10
• 作者: Xuying Ning, Katherine Tieu, Dongqi Fu

• 论文链接: https://huggingface.co/papers/2605.18747

Lance: Unified Multimodal Modeling by Multi-Task Synergy

• 英文标题: Lance: Unified Multimodal Modeling by Multi-Task Synergy
• 来源: 🤗 HuggingFace ⭐ 5/10
• 作者: Fengyi Fu, Mengqi Huang, Shaojin Wu

• 论文链接: https://huggingface.co/papers/2605.18678

KVPO: ODE-Native GRPO for Autoregressive Video Alignment via KV Semantic Exploration

• 英文标题: KVPO: ODE-Native GRPO for Autoregressive Video Alignment via KV Semantic Exploration
• 来源: 🤗 HuggingFace ⭐ 5/10
• 作者: Ruicheng Zhang, Kaixi Cong, Jun Zhou

• 论文链接: https://huggingface.co/papers/2605.14278

  📅 2026-05-19

Auditing Agent Harness Safety

• 英文标题: Auditing Agent Harness Safety
• 来源: 🤗 HuggingFace ⭐ 5/10
• 作者: Chengzhi Liu, Yichen Guo, Yepeng Liu

• 论文链接: https://huggingface.co/papers/2605.14271

MetaAgent-X : Breaking the Ceiling of Automatic Multi-Agent Systems via End-to-End Reinforcement Learning

• 英文标题: MetaAgent-X : Breaking the Ceiling of Automatic Multi-Agent Systems via End-to-End Reinforcement Learning
• 来源: 🤗 HuggingFace ⭐ 5/10
• 作者: Yaolun Zhang, Yujie Zhao, Nan Wang

• 论文链接: https://huggingface.co/papers/2605.14212

CM-EVS: Sparse Panoramic RGB-D-Pose Data for Complete Scene Coverage

• 英文标题: CM-EVS: Sparse Panoramic RGB-D-Pose Data for Complete Scene Coverage
• 来源: 🤗 HuggingFace ⭐ 5/10
• 作者: Jiale Liu, Jungang Li, Jieming Yu

• 论文链接: https://huggingface.co/papers/2605.15597

PhysBrain 1.0 Technical Report

• 英文标题: PhysBrain 1.0 Technical Report
• 来源: 🤗 HuggingFace ⭐ 5/10
• 作者: Shijie Lian, Bin Yu, Xiaopeng Lin

• 论文链接: https://huggingface.co/papers/2605.15298

From Plans to Pixels: Learning to Plan and Orchestrate for Open-Ended Image Editing

• 英文标题: From Plans to Pixels: Learning to Plan and Orchestrate for Open-Ended Image Editing
• 来源: 🤗 HuggingFace ⭐ 5/10
• 作者: Anirudh Sundara Rajan, Krishna Kumar Singh, Yong Jae Lee

• 论文链接: https://huggingface.co/papers/2605.15181

Learning from Failures: Correction-Oriented Policy Optimization with Verifiable Rewards

• 英文标题: Learning from Failures: Correction-Oriented Policy Optimization with Verifiable Rewards
• 来源: 🤗 HuggingFace ⭐ 5/10
• 作者: Mengjie Ren, Jie Lou, Boxi Cao

• 论文链接: https://huggingface.co/papers/2605.14539

Nudging Beyond the Comfort Zone: Efficient Strategy-Guided Exploration for RLVR

• 英文标题: Nudging Beyond the Comfort Zone: Efficient Strategy-Guided Exploration for RLVR
• 来源: 🤗 HuggingFace ⭐ 5/10
• 作者: Chanuk Lee, Sangwoo Park, Minki Kang

• 论文链接: https://huggingface.co/papers/2605.15726

MMSkills: Towards Multimodal Skills for General Visual Agents

• 英文标题: MMSkills: Towards Multimodal Skills for General Visual Agents
• 来源: 🤗 HuggingFace ⭐ 5/10
• 作者: Kangning Zhang, Shuai Shao, Qingyao Li

• 论文链接: https://huggingface.co/papers/2605.13527

FFAvatar: Few-Shot, Feed-Forward, and Generalizable Avatar Reconstruction

• 英文标题: FFAvatar: Few-Shot, Feed-Forward, and Generalizable Avatar Reconstruction
• 来源: 🤗 HuggingFace ⭐ 5/10
• 作者: Thuan Hoang Nguyen, Jiahao Luo, Yinyu Nie

• 论文链接: https://huggingface.co/papers/2605.15320

Flash-GRPO: Efficient Alignment for Video Diffusion via One-Step Policy Optimization

• 英文标题: Flash-GRPO: Efficient Alignment for Video Diffusion via One-Step Policy Optimization
• 来源: 🤗 HuggingFace ⭐ 5/10
• 作者: Xiaoxuan He, Siming Fu, Zeyue Xue

• 论文链接: https://huggingface.co/papers/2605.15980

HodgeCover: Higher-Order Topological Coverage Drives Compression of Sparse Mixture-of-Experts

• 英文标题: HodgeCover: Higher-Order Topological Coverage Drives Compression of Sparse Mixture-of-Experts
• 来源: 🤗 HuggingFace ⭐ 5/10
• 作者: Tao Zhong, Dongzhe Zheng, Christine Allen-Blanchette

• 论文链接: https://huggingface.co/papers/2605.13997

  📅 2026-05-18

今日无新增 OR 相关研究

📅 2026-05-17

今日无新增 OR 相关研究

📅 2026-05-17

今日无新增 OR 相关研究

📅 2026-05-17

今日无新增 OR 相关研究

📅 2026-05-17

今日无新增 OR 相关研究

📅 2026-05-16

根据强制筛选原则，对提供的论文列表进行分析后，筛选出属于“AI + 运筹优化”交叉领域的高质量研究如下：

---

大规模多智能体路径规划中的局部通信学习

• 英文标题: Learning to Communicate Locally for Large-Scale Multi-Agent Pathfinding
• 作者: 未在摘要中明确列出主要作者
• 核心贡献: 提出了一种基于机器学习的、可扩展的、去中心化的多智能体路径规划求解器，通过局部通信解决大规模MAPF问题。
• 实践价值: 可应用于物流仓储中的多机器人协同、搜救任务中的多智能体协调等大规模、实时性要求高的场景。
• OR 技术关键词: 多智能体路径规划 (MAPF), 强化学习, 路径规划 (PATHFINDING), 调度 (SCHEDULING)

• 论文链接: https://huggingface.co/papers/2605.07637

  超越单一智能：基于LLM的多智能体系统中的协作、失败归因与自我进化研究综述

• 英文标题: Beyond Individual Intelligence: Surveying Collaboration, Failure Attribution, and Self-Evolution in LLM-based Multi-Agent Systems
• 作者: 未在摘要中明确列出主要作者
• 核心贡献: 系统性综述了基于大语言模型的多智能体系统中的协作机制、故障诊断与传播，以及系统自我进化的挑战与方法。
• 实践价值: 为设计更可靠、更具鲁棒性的复杂自动化系统和决策支持系统提供理论框架，例如在供应链协调、智能制造产线管理等场景下，多智能体系统的鲁棒性至关重要。
• OR 技术关键词: 多智能体系统, 鲁棒性 (Robustness), 决策支持 (Decision Support)

• 论文链接: https://huggingface.co/papers/2605.14892

  SPIN：面向工业任务的结构化LLM规划与迭代导航

• 英文标题: SPIN: Structural LLM Planning via Iterative Navigation for Industrial Tasks
• 作者: 未在摘要中明确列出主要作者
• 核心贡献: 提出SPIN框架，一种结合了有向无环图（DAG）验证和前缀式执行控制的规划方法，确保LLM生成的工业工作流程结构有效且长度最优。
• 实践价值: 适用于工业自动化场景中的任务规划与执行，例如生产调度（PRODUCTION SCHEDULING）、物流流程编排等，通过减少无效规划和API调用成本来提升效率。
• OR 技术关键词: 规划 (PLANNING), 调度 (SCHEDULING), 生产 (PRODUCTION), 图优化 (DAG)

• 论文链接: https://huggingface.co/papers/2605.14051

  Nexus：面向时间序列预测的智能体框架

• 英文标题: Nexus: An Agentic Framework for Time Series Forecasting
• 作者: 未在摘要中明确列出主要作者
• 核心贡献: 提出了一个多智能体框架Nexus，通过融合时间序列基础模型（数值模式专家）和LLM（非结构化文本推理专家），解决了跨领域时间序列预测对上下文信号的需求。
• 实践价值: 可直接应用于销售预测（DEMAND FORECAST）、库存管理（INVENTORY MANAGEMENT）、供应链风险预测（SUPPLY CHAIN RISK FORECAST）等需要结合数值和事件信息的决策场景。
• OR 技术关键词: 预测 (FORECAST), 供应链 (SUPPLY CHAIN), 库存 (INVENTORY), 决策制定 (DECISION MAKING), 多智能体系统

• 论文链接: https://huggingface.co/papers/2605.14389

  动态潜路由

• 英文标题: Dynamic Latent Routing
• 作者: 未在摘要中明确列出主要作者
• 核心贡献: 提出了广义Dijkstra搜索（GDS）算法，并基于其原理设计了动态潜路由（DLR）方法，将马尔可夫决策过程（MDP）中的最优策略分解并组合为子策略，同时学习离散潜代码以实现高效路由。
• 实践价值: 可应用于需要战略规划与路径选择的复杂决策问题，例如物流网络中的动态路由（ROUTE OPTIMIZATION）、调度（SCHEDULING）问题的求解。
• OR 技术关键词: 路径规划 (ROUTE), 调度 (SCHEDULING), 马尔可夫决策过程 (MDP), 决策制定 (DECISION MAKING)

• 论文链接: https://huggingface.co/papers/2605.14323