arXiv 论文速递

推理模型生成冗长思维链导致推理成本高昂，该工作用自监督置信度训练让模型学会提前停止，无需额外停止标签即可显著提升效率。

Reasoning models often generate very long reasoning traces, making inference computationally expensive. Existing approaches typically improve efficiency either through inference-ti…

差分隐私PCA此前依赖特征值间隙假设，该工作给出无间隙条件的算法，使隐私保护主成分分析在更广泛数据上可用。

We give a gap-free differentially private algorithm for the principal component analysis (PCA) problem with Gaussian data.

Recent literature has shown a strong connection between optimization and sampling. We develop the corresponding first-order theory for diffusion models. First, the SDE-based revers…

Generative AI systems are increasingly used, but aligning their outputs with user requirements poses a continuing challenge. Here, we aim to ensure that the distribution of an attr…

大模型隐式推断用户属性却难以被审计，论文用信念自蒸馏提取用户模型，为LLM用户建模的可解释性与隐私评估提供新手段。

Large language models (LLMs) implicitly infer attributes of their users and adapt their behavior accordingly, yet these beliefs remain difficult to inspect and causally manipulate.…

多个独立训练的LoRA适配器合并时会相互干扰，论文提出新技能应只读不写的原则，为低成本多任务微调的可靠组合提供关键指导。

Low-rank adapters (LoRA) make it cheap to fine-tune a large language model once per task, but combining several independently trained adapters into one model remains difficult: mer…

Direct feedback alignment (DFA) trains hidden layers through fixed random projections of output error. With tanh hidden units and independent sigmoid outputs, plain stochastic grad…

We investigate whether natural-language documentation helps coding agents resolve software issues, and we build the tools to construct and evaluate it. We introduce a roundtrip ben…

决策Transformer在长程 rollout 中因条件上下文漂移而性能下降，该工作引入信任引导缓解分布外漂移，提升序列决策的长期稳定性。

Decision Transformer performance degrades on long rollouts because the conditioning context drifts out of the training distribution. We show that this drift is visible through the …

Uneven rotation and dropped frames make equal-angle assumptions unreliable for turntable reconstruction. We present OC-GS, an object-centric Gaussian splatting that refines each im…

Deep learning has substantially accelerated the calibration of complex stochastic-volatility models, but neural point calibration alone does not capture the uncertainty remaining a…

Conversational AI tools are entering children's everyday experiences, and schools are interested in adopting them. However, successful classroom integration depends not only on the…