AI资讯日报 - 2026/7/28

👨‍🔬 Hangjie Yuan, Yichen Qian, Zhiwei Tang, Xianzhe Xu, Lirong Wu, Sicheng Yang, Jinwang Wang, Pengju Wang, Zhitao Zeng, Yizeng Han, Yan Xing, Shengxuan Luo, Tao Feng, Qing Xie, Weigen Yao, Yi Yang, Zuozhu Liu, Jiasheng Tang, Shaocheng Wang, Jitao Wang, Jiahong Dong, Weihua Chen, Feng Xu, Fan Wang

Rethinking Classifier-Free Guidance in On-Policy Diffusion Distillation

学术论文 ArXiv 重要度: 8

发现扩散模型在线策略蒸馏中的负分支不对称问题，提出正方向匹配（PDM）目标，提升视频控制蒸馏鲁棒性。

👨‍🔬 Bingnan Li, Haozhe Wang, Haozhong Xiong, Fangtai Wu, Jinpeng Yu, Yang Shi, Jiaming Liu, Ruihua Huang

KANEx: Translating Kolmogorov-Arnold Networks' Interpretability to Medical Explainability

学术论文 ArXiv 重要度: 8

利用KAN网络的符号可解释性，首次将内在透明度用于VLM推理，生成更可信的医学解释和热力图。

👨‍🔬 Krithi Shailya, Ananya Lakshmi Ravi, Venkatanathan K. V., Sowmya S. Sundaram, Gokul S. Krishnan, Aditi Anand, Balaraman Ravindran

The Physics of Multi-Turn Long-Horizon Planning: From Pre-training to Post-training via Single- and Multi-Teacher On-Policy Agentic Distillation

学术论文 ArXiv 重要度: 8

系统研究多轮长程规划能力在预训练和后训练中的获取、塑造与整合，提出多教师在线策略蒸馏方法。

👨‍🔬 Tianyi Men, Zhuoran Jin, Kang Liu, Jun Zhao

DataOrchestra: Learning to Orchestrate Per-Example Curation of Pretraining Data

学术论文 ArXiv 重要度: 7

提出DataOrchestra框架，为每个预训练样本动态编排处理流水线，提升下游性能并降低计算开销。

👨‍🔬 Zhen Huang, Yikun Wang, Shijie Xia, Pengfei Liu

Efficient LLM-Generated Shuttling Compilers for Complex Trapped-Ion Architectures

学术论文 ArXiv 重要度: 7

首次使用LLM自动生成离子阱量子计算机的穿梭编译器，性能优于手工编写，开发时间从数月缩短至数天。

👨‍🔬 Fabian Kreppel, Reza Salkhordeh, Ferdinand Schmidt-Kaler, André Brinkmann

Denial of Deadline: Network-Driven Accuracy Collapse in Distributed Inference Pipelines

学术论文 ArXiv 重要度: 7

发现分布式推理管道中，网络攻击可导致慢路径结果超时丢弃，造成精度崩溃，影响自动驾驶等应用。

👨‍🔬 Jhonatan Tavori, Gur-Eyal Sela, Ion Stoica, Gil Zussman

A corrective agentic hybrid RAG and an operations-grounded evaluation for a scientific facility

学术论文 ArXiv 重要度: 7

为先进光子源部署APS-RAG系统，融合多种检索与纠错机制，在设施运营知识问答上取得改进。

👨‍🔬 Rajat Sainju, Dariusz Jarosz, Hairong Shang, Michael Prince, Ryan M. Aydelott, Mathew J. Cherukara, Yine Sun, Michael D. Borland

ERUnderstand: Evaluating Vision-Language Models on Structured ER Diagrams

学术论文 ArXiv 重要度: 6

提出ER图理解基准ERUnderstand，评估VLM在数据库概念设计中的结构化理解能力，发现弱实体和N元关系性能差。

👨‍🔬 Ali Ansari, Yasmin Mohammadi, Farnoush Nili, Parsa Esmaeilkhani, Longin Jan Latecki, Eduard Dragut

Co-Learning for Missing Arbitrary Modalities in Multi-modal Classification

学术论文 ArXiv 重要度: 6

提出多模态协同学习框架处理任意模态缺失，在特征和决策层面利用信息，提升鲁棒性。

👨‍🔬 Francisco Mena, Dino Ienco, Roberto Interdonato, Cassio F. Dantas, Simon Besnard

Eviction as Estimation: A Fixed-Lag Smoothing View of Test-Time Memory, and When Measuring Beats Accumulating

学术论文 ArXiv 重要度: 6

将语言模型记忆驱逐重新定义为固定滞后平滑估计问题，提出基于效用的方法，但在标准基准上优势有限。

👨‍🔬 Maruthi Vemula, Neeraj Praneeth Gajula

Reason-Mediated Behavioral Models for Auditing LLM Social Simulators

学术论文 ArXiv 重要度: 6

提出通过理由状态审计LLM社会模拟器，发现模拟理由常与人类证据不符，仅匹配最终答案不足够。

👨‍🔬 Atharva Pandey, Gautam Jajoo

🤖 AI资讯日报

📊 今日趋势总结

MIT Non-AI License

Why Boring Businesses Outlast AI Hype Cycles

Ask HN: Anyone concerned about NYC Local Law 144?

Ask HN: Is the rate of progress in AI exponential?

The AI Crackpot Index

Ask HN: What's the pain using current AI algorithms?

Common Lisp + Machine Learning Internship at Google (Mountain View, CA)

Ask HN: What would you read to learn about "artificial intelligence"?

Show HN: Startup Raising capital through Book Sales

The Next Bill Gates or Albert Einstein in AI “Chris Clark” – Yourobot

Bioinformatician

NLP, AI, ML, bots – a passing trend or much more? What's your take on this?

ClinFusion: A Vision-Centric Multimodal LLM System for Holistic Medical Understanding

Rethinking Classifier-Free Guidance in On-Policy Diffusion Distillation

KANEx: Translating Kolmogorov-Arnold Networks' Interpretability to Medical Explainability

The Physics of Multi-Turn Long-Horizon Planning: From Pre-training to Post-training via Single- and Multi-Teacher On-Policy Agentic Distillation

DataOrchestra: Learning to Orchestrate Per-Example Curation of Pretraining Data

Efficient LLM-Generated Shuttling Compilers for Complex Trapped-Ion Architectures

Denial of Deadline: Network-Driven Accuracy Collapse in Distributed Inference Pipelines

A corrective agentic hybrid RAG and an operations-grounded evaluation for a scientific facility

ERUnderstand: Evaluating Vision-Language Models on Structured ER Diagrams

Co-Learning for Missing Arbitrary Modalities in Multi-modal Classification

Eviction as Estimation: A Fixed-Lag Smoothing View of Test-Time Memory, and When Measuring Beats Accumulating

Reason-Mediated Behavioral Models for Auditing LLM Social Simulators

📅 历史日报目录