arXiv 论文速递
· 2026-09-23
研究表示自编码器高维潜空间的扩散可行性，揭示预训练视觉编码器特征空间用于扩散生成的条件与限制，对生成模型架构设计有指导意义。
Representation Autoencoders (RAEs) enable diffusion models to operate in the feature spaces of pretrained visual encoders. However, many off-the-shelf encoders are not optimized fo…
· 2026-09-23
Our results show that contrastive learning outperforms a classification-based approach to authorship verification under the tested settings. We identify loss function, batch size, …
通过对照实验证明AI辅导与人类辅导在GRE学习收益上相当，为AI教育规模化应用提供了少见的严格实证证据，对教育公平与普及意义重大。
Artificial intelligence offers an unprecedented opportunity to augment human capabilities, yet progress at the frontier has focused primarily on advancing model capabilities. We in…
Social navigation typically assumes a specified goal and focuses on reaching it while respecting social conventions, whereas robot group joining requires predicting where to join b…
We introduce Sharper Transductive Local Complexity (STLC), a localized complexity method for transductive learning under uniform sampling without replacement. The construction star…
现有代码基准多测静态理解，本文构建仓库级动态基准考察LLM对运行时行为的推理能力，直击代码智能体的关键短板，对评估与改进LLM编程能力有重要价值。
Large language models (LLMs) are increasingly used in coding tasks, but their ability to reason about code execution remains unclear. Existing repository-level QA benchmarks mainly…
· 2026-09-23
We study the nonequilibrium dynamics of a minimal recurrent transformer with $N$ normalized tokens, $Q=K=I$, and a negative value map $V=-I$. Similarity-based attention selects nea…
发现数学推理模型答案对规则顺序不变但内部表征敏感，揭示推理鲁棒性与表征结构的脱节，对理解和改进LLM推理机制有启发价值。
Reordering a set of mathematical rules without changing its meaning should preserve the correct answer, but must a model's internal representations stay invariant too? We investiga…
· 2026-09-23
We study minimal-norm interpolation and $\ell_2$-regularized logistic-loss minimization for binary classification by univariate two-layer ReLU networks. We give complete geometric …
Personalizing exoskeleton assistance across operating conditions is constrained by the time and physical effort required to collect user feedback. We examined whether a user's pref…
Recursive state estimation often executes approximate numerical solutions inside a feedback loop, where highly accurate local steps do not guarantee better overall results. For a f…
提出面向LLM智能体的世界模型新范式，用编辑式建模替代传统预测式建模，有望提升长程任务规划效率，是智能体研究的重要方向探索。
Recent advances in large language models (LLMs) have enabled agents to tackle long-horizon tasks across diverse environments. To further improve agent performance, existing languag…