AI资讯日报 - 2026/9/11

👨‍🔬 Yunfei Ge, Anbang Liu, Qineng Wang, Johnalbert Garnica, Jianwen Lyu, Zihan Wang, Reuben Tan, Jianfeng Gao, Ruohan Zhang, Yining Hong, Jiajun Wu, Manling Li

Artificial Id: Drive and Persistent Alignment in Agentic AI

学术论文 ArXiv 重要度: 9

提出“人工本我”作为智能体的自适应内在驱动力，实验证明无需显式目标即可涌现持续行为，但需建立持久对齐边界以防状态污染与错位。

👨‍🔬 Yakov Pyotr Shkolnikov

Can Edge-Deployable Vision-Language Models Identify Species?

学术论文 ArXiv 重要度: 8

评估2-8B边缘VLM在物种识别任务上的表现，所有模型在野外相机陷阱图像上性能骤降，专业小模型BioCLIP仍大幅领先，且存在编造物种名现象。

👨‍🔬 William Zhou, Mayukha Siripuram, Xiao Yan, Ziqi Liu, Yi Ding

Domain-Specific Hallucination Detection in Large Language Models

学术论文 ArXiv 重要度: 8

结合DeBERTa分类、MC Dropout与温度校准的幻觉检测流水线在HaluEval上F1达0.915，并用DPO将生成器幻觉率从85.5%降至37.7%，跨域迁移需领域微调。

👨‍🔬 Varun Teja Chundru, Debasmita Biswas

General Quantification of Covariate and Concept Shifts

学术论文 ArXiv 重要度: 8

利用熵最优传输提出γ*-概念偏移，统一协变量与概念偏移的误差界，并给出可估计的DataShifts算法，为分布偏移下的学习误差提供严谨量化工具。

👨‍🔬 Hongbo Chen, Li Charlie Xia

Biology-in-the-loop: Amortized Adaptive Hit Discovery in CRISPR Screens

学术论文 ArXiv 重要度: 7

推出AssayBench-Loop基准与AssayLoop框架，结合Transformer摊销策略与LLM先验，在仅检测5%候选库时回收27.7%命中，较随机选择富集5.67倍。

👨‍🔬 Carl Edwards, Edward De Brouwer, Xiner Li, Namkyeong Lee, Ehsan Hajiramezanali, Anne Biton, Sara Mostafavi, Gabriele Scalia

The Last AI Built by Humans: Toward Genuine Recursive Self-Improvement

学术论文 ArXiv 重要度: 7

提出递归自我改进（RSI）发展路线图，从执行自主到元改进，并分析科学发现、具身智能等场景的差异化需求与关键挑战。

👨‍🔬 Yi Duan, Ying Liu, Zirui Tang, Haodong Chen, Jun Zhou, Yumou Liu, Bangrui Xu, Yukai Wu, Sidi Chen, Yuhan Zhou, Haoyu Wang, Xiaoyou Yu, Shaokun Han, Xuzhou Zhu, Le Zhou, Bolin Lu, Wei Zhou, Jiachen Liu, Nuozhou Fang, Jiaxin Tian, Ruoyu Chen, Yuxuan Li, Kai Zuo, Kaiyan Zhang, Jiantao Qiu, Conghui He, Guoliang Li, Bowen Zhou, Zhiyuan Liu, Zhoufutu Wen, Jihua Kang, Xuanhe Zhou, Fan Wu

RetroThinker: Enabling Retrospective Thinking in Speech LLMs

学术论文 ArXiv 重要度: 7

为流式语音大模型Moshi引入回顾式思维链，通过SFT与长度DPO优化，在GSM8K上以相近延迟实现11%绝对准确率提升。

👨‍🔬 Yi-Jen Shih, Puyuan Peng, Abdelrahman Mohamed, David Harwath

Generative Marketing Mix Modeling: A Causal Inference Framework Linking GEO and GEM to Business Impact

学术论文 ArXiv 重要度: 6

提出生成式营销组合模型GMMM，融合生成答案曝光与赞助投放记录，建立GEO/GEM因果效应识别条件，并利用英日双语模拟数据验证。

👨‍🔬 Masahiro Kato, Daiki Honma, Taka Kato

On the Regularization Landscape for the Linear Recommendation Models

学术论文 ArXiv 重要度: 6

发现线性推荐模型性能领先者仅依赖核范数或Frobenius范数正则，据此提出两种兼具低秩、闭式解与表达力的新方案，统一现有方法。

👨‍🔬 Dong Li, Zhenming Liu, Ruoming Jin, Hao Zhou, Zhi Liu, Jing Gao, Bin Ren

Explainability Assistant: A Conversational XAI Interface for Interpreting Energy Consumption Models

学术论文 ArXiv 重要度: 6

开源对话式XAI系统Explainability Assistant利用LLM函数调用，意图解析准确率达94%，能源领域专家一致偏好其相对传统仪表盘的易用性。

👨‍🔬 Rodion Krjutškov, Eduard Barbu, Nikos Sakkas, Sofia Yfanti

🤖 AI资讯日报

📊 今日趋势总结

MIT Non-AI License

Ask HN: Anyone concerned about NYC Local Law 144?

Why Boring Businesses Outlast AI Hype Cycles

The AI Crackpot Index

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

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

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

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

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

Show HN: Startup Raising capital through Book Sales

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

Bioinformatician

GPU-CFR: 80x Faster Counterfactual Regret Minimization by Compiling the Game to Static Dataflow and CUDA Graph Replay

MindTopo: Can Foundation Models Reason in Topological Space?

Artificial Id: Drive and Persistent Alignment in Agentic AI

Can Edge-Deployable Vision-Language Models Identify Species?

Domain-Specific Hallucination Detection in Large Language Models

General Quantification of Covariate and Concept Shifts

Biology-in-the-loop: Amortized Adaptive Hit Discovery in CRISPR Screens

The Last AI Built by Humans: Toward Genuine Recursive Self-Improvement

RetroThinker: Enabling Retrospective Thinking in Speech LLMs

Generative Marketing Mix Modeling: A Causal Inference Framework Linking GEO and GEM to Business Impact

On the Regularization Landscape for the Linear Recommendation Models

Explainability Assistant: A Conversational XAI Interface for Interpreting Energy Consumption Models

📅 历史日报目录