AI资讯日报 - 2026/8/2

👨‍🔬 Xiangning Lin, Shenzhe Zhu, Shu Yang, Zhenyu Zhang, Haoqian Zhang, Yipeng Zhao, Chengxuan Qian, Tianwei Wang, Ziheng Zhang, Zhenlong Yuan, Dingcheng Wang, Juncheng Wu, Yuan Si, Jiaxin Liu, Baolong Bi, Robert Mahari, Tobin South, Dazza Greenwood, Zexue He, Rishi Bommasani, Sophia Kazinnik, Andreas Haupt, Samuele Marro, Erik Brynjolfsson, Alex Pentland, Jiaxin Pei

OSReward: Instituting Standardized Evaluation for Cross-Platform Computer-Use Reward Models

学术论文 ArXiv 重要度: 9

发布OSReward基准，评估VLM作为计算机使用智能体轨迹评判的可靠性，发现现有模型存在系统性宽容偏差，并训练开源奖励模型OS-Shepherd。

👨‍🔬 Qiushi Sun, Kanzhi Cheng, Yian Wang, Bowen Yang, Hang Yan, Liheng Chen, Fangzhi Xu, Zichen Ding, Nuo Chen, Jialin Cao, Xingdong Gong, Zehao Li, Kaiming Jin, Xinfeng Yuan, Zhoumianze Liu, Jingyang Gong, Zhangyue Yin, Jiahui Gao, Zhiyong Wu, Tianbao Xie, Jianbing Zhang, Ben Kao, Lingpeng Kong

Sample More, Reflect Less: Self-Refine and Reflexion Lose to Repeated Sampling at Equal Token Cost, from 1.5B to 7B

学术论文 ArXiv 重要度: 9

实验证明在相同token预算下，重复采样并投票优于自省方法（如Self-Refine），且自省方法在小型模型上表现更差，挑战其有效性。

👨‍🔬 Iliya Mirzaei

APO: Unsupervised Atomic Policy Optimization for 3D Structure Prediction of Atomic Systems

学术论文 ArXiv 重要度: 9

提出无监督原子策略优化框架APO，利用双重奖励机制无需真实结构标签，在晶体和抗体结构预测上超越监督基线，并提升推理效率。

👨‍🔬 Shentong Mo, Yatao Bian

Learning to Trace Seiberg Dualities

学术论文 ArXiv 重要度: 8

利用机器学习方法建立超对称quiver规范理论的Seiberg对偶，发现transformer和MLP优于确定性算法，结合路径算法可提升效率。

👨‍🔬 Jonathan J. Heckman, Shani Meynet, Alessandro Mininno, Gary Shiu

PAC-MAN: Perception-Aware CBF-RL for Whole-Body Safety in Humanoid Dodgeball

学术论文 ArXiv 重要度: 8

结合控制屏障与强化学习，实现人形机器人躲避球任务的全身安全控制，在真实机器人上达到95%成功率，验证了感知与安全约束的集成。

👨‍🔬 Lizhi Yang, Junheng Li, Aaron D. Ames

AskChem: Claim-Centered Infrastructure for Chemistry Literature Synthesis

学术论文 ArXiv 重要度: 8

构建以声明为中心的化学文献检索基础设施，索引240万声明，提供多种检索方式，显著提升AI代理在化学文献综合中的准确性和可追溯性。

👨‍🔬 Bing Yan, Gregory Wolfe, Stefano Martiniani, Kyunghyun Cho

PAIChecker: Uncovering and Checking PR-Issue Misalignment in SWE-Bench-Like Benchmarks

学术论文 ArXiv 重要度: 8

发现SWE-bench中13.6%的PR-Issue对存在错位，提出多智能体系统PAIChecker自动检测，在多个基准上达到高准确率。

👨‍🔬 Manyi Wang, Junjielong Xu, Pinjia He

Rethinking Inference-Time Scaling in Local Computer-Use Agents: Failure Modes and Compute Tradeoffs

学术论文 ArXiv 重要度: 8

系统研究本地计算机使用智能体的推理时间扩展，发现额外计算收益递减且改变失败模式，提出需要选择性计算分配和失败感知控制。

👨‍🔬 Woongkyu Lee, Jungwook Choi

DualG-MRAG: Decoupling Macro-Reasoning and Micro-Matching for Multimodal Retrieval-Augmented Generation

学术论文 ArXiv 重要度: 7

提出双图框架解耦宏观推理和微观匹配，通过GNN检索器进行动态消息传递，提升多模态RAG在复杂多跳推理任务上的性能。

👨‍🔬 Jiacheng Tao, Qingyun Sun, Haonan Yuan, Ziwei Zhang, Jianxin Li

Algorithms for Structured Elections under Thiele Voting Rules

学术论文 ArXiv 重要度: 6

研究Thiele投票规则下的获胜者确定问题，在Voter Interval域上设计FPT算法，解决PAV的复杂度问题，并回答两个开放问题。

👨‍🔬 Alexandra Lassota, Krzysztof Sornat

🤖 AI资讯日报

📊 今日趋势总结

Why Boring Businesses Outlast AI Hype Cycles

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

Ask HN: Anyone concerned about NYC Local Law 144?

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

MIT Non-AI License

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

The AI Crackpot Index

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

Show HN: Startup Raising capital through Book Sales

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

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

Bioinformatician

ReToken: One Token to Improve Vision-Language Models for Visual Retrieval

AISPA: User-Centric System Prompt Auditing for Large Language Model Applications

OSReward: Instituting Standardized Evaluation for Cross-Platform Computer-Use Reward Models

Sample More, Reflect Less: Self-Refine and Reflexion Lose to Repeated Sampling at Equal Token Cost, from 1.5B to 7B

APO: Unsupervised Atomic Policy Optimization for 3D Structure Prediction of Atomic Systems

Learning to Trace Seiberg Dualities

PAC-MAN: Perception-Aware CBF-RL for Whole-Body Safety in Humanoid Dodgeball

AskChem: Claim-Centered Infrastructure for Chemistry Literature Synthesis

PAIChecker: Uncovering and Checking PR-Issue Misalignment in SWE-Bench-Like Benchmarks

Rethinking Inference-Time Scaling in Local Computer-Use Agents: Failure Modes and Compute Tradeoffs

DualG-MRAG: Decoupling Macro-Reasoning and Micro-Matching for Multimodal Retrieval-Augmented Generation

Algorithms for Structured Elections under Thiele Voting Rules

📅 历史日报目录