Recent Posts

The End of Coding: Andrej Karpathy on AI Agents, AutoResearch, and the Loopy Era

编程的终结：Andrej Karpathy 谈 AI 智能体、自动研究与循环时代

3/23/2026

AIAgentsAutoResearch

CubiD: Breaking the Dimensionality Ceiling in Discrete Visual Generation

CubiD：突破离散视觉生成的维度天花板

3/21/2026

arxivaics.cv

AI Agents Can Already Autonomously Perform Experimental High Energy Physics

AI 智能体已经可以自主执行实验高能物理学

3/20/2026

arxivaiagents

Do VLMs Actually Need Vision Transformers? A Case for SSM Encoders

视觉语言模型真的需要视觉Transformer吗？状态空间模型编码器的潜力

3/20/2026

arxivaics.cv

Evolving Jailbreaks: Automated Multi-Objective Long-Tail Attacks on LLMs

演化越狱：LLM 的自动化多目标长尾攻击

3/20/2026

arxivaisecurity

Experience is the Best Teacher: Motivating Effective Exploration in RL for LLMs

经验是最好的老师：激励 LLM 强化学习中的有效探索

3/20/2026

arxivaireinforcement-learning

Learning Dynamic Belief Graphs for Theory-of-mind Reasoning

学习动态信念图进行心智理论推理

3/20/2026

arxivaitheory-of-mind

Measuring Faithfulness Depends on How You Measure: Classifier Sensitivity in LLM CoT Evaluation

测量忠实度取决于如何测量：LLM 思维链评估中的分类器敏感性

3/20/2026

arxivaillm

Nemotron-Cascade 2: Teaching a Small Model to Think Big with CascadeRL and On-Policy Distillation

Nemotron-Cascade 2：用级联强化学习和在线蒸馏让小模型学会深度推理

3/20/2026

arxivaics.cl

Pitfalls in Evaluating Interpretability Agents

评估可解释性智能体的陷阱

3/20/2026

arxivaiinterpretability

R-Equivalence on Cubic Surfaces: Closing a50-Year Gap with AI-Assisted Profs

三次曲面上的R-等价：用AI辅助证明填补五十年空白

3/20/2026

arxivaimath.ag

Semantic Token Clustering for Efficient Uncertainty Quantification in LLMs

语义令牌聚类：LLM 中的高效不确定性量化

3/20/2026

arxivaillm