AI资讯日报 - 2025/12/11

👨‍🔬 Justin W. Lin, Eliot Krzysztof Jones, Donovan Julian Jasper, Ethan Jun-shen Ho, Anna Wu, Arnold Tianyi Yang, Neil Perry, Andy Zou, Matt Fredrikson, J. Zico Kolter, Percy Liang, Dan Boneh, Daniel E. Ho

LISN: Language-Instructed Social Navigation with VLM-based Controller Modulating

学术论文 ArXiv 重要度: 8

提出首个语言指令社会导航基准LISN-Bench及分层系统Social-Nav-Modulator，使机器人能遵循复杂指令进行社交导航，成功率显著提升。

👨‍🔬 Junting Chen, Yunchuan Li, Panfeng Jiang, Jiacheng Du, Zixuan Chen, Chenrui Tie, Jiajun Deng, Lin Shao

FlipLLM: Efficient Bit-Flip Attacks on Multimodal LLMs using Reinforcement Learning

学术论文 ArXiv 重要度: 8

提出FlipLLM框架，利用强化学习高效发现多模态大模型的比特翻转攻击漏洞，显著加速关键位识别并指导硬件防御。

👨‍🔬 Khurram Khalil, Khaza Anuarul Hoque

Efficient Continual Learning in Neural Machine Translation: A Low-Rank Adaptation Approach

学术论文 ArXiv 重要度: 7

采用低秩适应框架解决神经机器翻译的持续学习问题，实现参数高效微调、实时交互适应并缓解灾难性遗忘。

👨‍🔬 Salvador Carrión, Francisco Casacuberta

Provably Learning from Modern Language Models via Low Logit Rank

学术论文 ArXiv 重要度: 7

基于现代语言模型经验上具有低对数秩的观察，提出首个可证明从此类模型中学习的算法，为理解大模型提供了理论框架。

👨‍🔬 Noah Golowich, Allen Liu, Abhishek Shetty

STACHE: Local Black-Box Explanations for Reinforcement Learning Policies

学术论文 ArXiv 重要度: 7

提出STACHE框架，为强化学习策略提供局部黑盒解释，生成鲁棒性区域和最小反事实，揭示策略决策边界与演化。

👨‍🔬 Andrew Elashkin, Orna Grumberg

SCOPE: Language Models as One-Time Teacher for Hierarchical Planning in Text Environments

学术论文 ArXiv 重要度: 6

提出SCOPE方法，利用LLM生成子目标仅一次初始化轻量级学生模型，实现文本环境中高效分层规划，大幅降低推理时间。

👨‍🔬 Haoye Lu, Pavan Seshadri, Kaheer Suleman

Visual Heading Prediction for Autonomous Aerial Vehicles

学术论文 ArXiv 重要度: 6

提出视觉框架实现无人机对地面车辆的实时检测与航向角预测，在GPS拒止环境下支持可靠的多智能体协调。

👨‍🔬 Reza Ahmari, Ahmad Mohammadi, Vahid Hemmati, Mohammed Mynuddin, Parham Kebria, Mahmoud Nabil Mahmoud, Xiaohong Yuan, Abdollah Homaifar

FALCON: Few-step Accurate Likelihoods for Continuous Flows

学术论文 ArXiv 重要度: 6

提出FALCON方法，通过混合训练目标实现连续流模型的少步采样与精确似然计算，大幅提升分子玻尔兹曼采样的效率。

👨‍🔬 Danyal Rehman, Tara Akhound-Sadegh, Artem Gazizov, Yoshua Bengio, Alexander Tong

Supervised learning pays attention

学术论文 ArXiv 重要度: 6

将注意力机制引入监督学习，通过注意力加权为每个测试点拟合个性化局部模型，提升预测性能并保持可解释性。

👨‍🔬 Erin Craig, Robert Tibshirani

Human-in-the-Loop and AI: Crowdsourcing Metadata Vocabulary for Materials Science

学术论文 ArXiv 重要度: 5

提出MatSci-YAMZ平台，结合AI与人类参与（包括众包）开发材料科学元数据词汇表，验证了AI-人机协同模型的可行性。

👨‍🔬 Jane Greenberg, Scott McClellan, Addy Ireland, Robert Sammarco, Colton Gerber, Christopher B. Rauch, Mat Kelly, John Kunze, Yuan An, Eric Toberer

Bayesian Networks, Markov Networks, Moralisation, Triangulation: a Categorical Perspective

学术论文 ArXiv 重要度: 5

提出范畴论框架，将贝叶斯网络与马尔可夫网络之间的道德化与三角化转换建模为函子，为概率图模型理论引入函子视角。

👨‍🔬 Antonio Lorenzin, Fabio Zanasi

🤖 AI资讯日报

📊 今日趋势总结

Why Boring Businesses Outlast AI Hype Cycles

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: Anyone concerned about NYC Local Law 144?

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

The AI Crackpot Index

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

Ask HN: Dipping my toes with artificial intelligence and what to expect? (CS)

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

Bioinformatician

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

Show HN: Startup Raising capital through Book Sales

Comparing AI Agents to Cybersecurity Professionals in Real-World Penetration Testing

LISN: Language-Instructed Social Navigation with VLM-based Controller Modulating

FlipLLM: Efficient Bit-Flip Attacks on Multimodal LLMs using Reinforcement Learning

Efficient Continual Learning in Neural Machine Translation: A Low-Rank Adaptation Approach

Provably Learning from Modern Language Models via Low Logit Rank

STACHE: Local Black-Box Explanations for Reinforcement Learning Policies

SCOPE: Language Models as One-Time Teacher for Hierarchical Planning in Text Environments

Visual Heading Prediction for Autonomous Aerial Vehicles

FALCON: Few-step Accurate Likelihoods for Continuous Flows

Supervised learning pays attention

Human-in-the-Loop and AI: Crowdsourcing Metadata Vocabulary for Materials Science

Bayesian Networks, Markov Networks, Moralisation, Triangulation: a Categorical Perspective

📅 历史日报目录