AAAI

Capacity Constraints Make Admissions Processes Less Predictable

容量限制使录取过程更不具有可预测性

Evan Dong, Nikhil Garg, Sarah Dean

AI总结 容量限制使录取过程更不可预测，研究揭示了机器学习在录取决策中因容量约束导致的预测性能下降问题。

Comments This paper was accepted to the 2026 AAAI Conference on Artificial Intelligence AI for Social Impact Track

Let the Void Be Void: Robust Open-Set Semi-Supervised Learning via Selective Non-Alignment

让虚空成为虚空：通过选择性非对齐实现鲁棒的开集半监督学习

You Rim Choi, Subeom Park, Seojun Heo, Eunchung Noh, Hyung-Sin Kim

AI总结 SkipAlign通过引入选择性非对齐机制，有效区分分布内和分布外样本，提升开集半监督学习中对未知分布外实例的检测能力，同时保持分布内分类精度。

Comments Proceedings of the 40th AAAI Conference on Artificial Intelligence (AAAI 2026)

Optimized Algorithms for Text Clustering with LLM-Generated Constraints

基于LLM生成约束的文本聚类优化算法

Chaoqi Jia, Weihong Wu, Longkun Guo, Zhigang Lu, Chao Chen, Kok-Leong Ong

AI总结 本文提出了一种基于LLM生成约束的文本聚类优化算法，通过生成约束集提高效率和准确性，减少LLM查询次数20倍以上。

Comments AAAI-26

Graph Smoothing for Enhanced Local Geometry Learning in Point Cloud Analysis

图平滑用于点云分析中的增强局部几何学习

Shangbo Yuan, Jie Xu, Ping Hu, Xiaofeng Zhu, Na Zhao

AI总结 本文提出一种结合图平滑和局部几何学习的方法，以提升点云分析中边界点和连接区域的处理效果。

Comments Accepted by AAAI 2026

Finding the Translation Switch: Discovering and Exploiting the Task-Initiation Features in LLMs

寻找翻译切换：在大语言模型中发现并利用任务启动特征

Xinwei Wu, Heng Liu, Xiaohu Zhao, Yuqi Ren, Linlong Xu, Longyue Wang, Deyi Xiong, Weihua Luo, Kaifu Zhang

AI总结 本研究通过发现大语言模型中的翻译启动特征，揭示其内在翻译能力，并提出基于机理困难样本的高效微调策略，提升模型鲁棒性和效率。

Comments Accepted by AAAI 2026

PatientVLM Meets DocVLM: Pre-Consultation Dialogue Between Vision-Language Models for Efficient Diagnosis

PatientVLM 与 DocVLM 相遇：为高效诊断的预咨询对话

K Lokesh, Abhirama Subramanyam Penamakuri, Uday Agarwal, Apoorva Challa, Shreya K Gowda, Somesh Gupta, Anand Mishra

AI总结 本文提出预咨询对话框架，通过视觉语言模型模拟医生与患者对话，利用合成症状提升诊断效率。

Comments Accepted at AAAI 2026 Main Track

Realistic Curriculum Reinforcement Learning for Autonomous and Sustainable Marine Vessel Navigation

现实课程强化学习用于自主可持续海洋船舶导航

Zhang Xiaocai, Xiao Zhe, Liang Maohan, Liu Tao, Li Haijiang, Zhang Wenbin

AI总结 本文提出基于现实课程强化学习的框架，结合数据驱动的海洋模拟和机器学习预测模块，以实现自主且可持续的船舶导航。

Comments Present in The 40th Annual AAAI Conference on Artificial Intelligence (AAAI-26)

Zero-Shot Transfer Capabilities of the Sundial Foundation Model for Leaf Area Index Forecasting

Sundial基础模型在叶面积指数预测中的零样本转移能力

Peining Zhang, Hongchen Qin, Haochen Zhang, Ziqi Guo, Guiling Wang, Jinbo Bi

机构 * Sundial Foundation（Sundial基金会） 纠错

AI总结 Sundial基础模型在无需任务特定调优的情况下，能够超越专门监督模型，在遥感时间序列预测中展现强大零样本转移能力。

Comments 6 pages, 5 figures, AAAI 2026 AgriAI workshop

Causal Inference under Threshold Manipulation: Bayesian Mixture Modeling and Heterogeneous Treatment Effects

阈值操纵下的因果推断：贝叶斯混合建模与异质性处理效应

Kohsuke Kubota, Shonosuke Sugasawa

AI总结 本文提出了一种贝叶斯混合建模方法，用于在阈值操纵下估计因果效应，并扩展至分层贝叶斯框架以估计异质性处理效应。

Comments Paper accepted to AAAI 2026

Hearing More with Less: Multi-Modal Retrieval-and-Selection Augmented Conversational LLM-Based ASR

听更清晰，用更少：多模态检索与选择增强的对话式LLM基于ASR

Bingshen Mu, Hexin Liu, Hongfei Xue, Kun Wei, Lei Xie

AI总结 本文提出MARS方法，通过多模态检索与选择提升对话式LLM-ASR的准确性，仅用1.5K小时数据即可超越训练于179K小时数据的顶级系统。

Comments AAAI 2026

Lost in Benchmarks? Rethinking Large Language Model Benchmarking with Item Response Theory

陷入基准？基于项目反应理论重新思考大语言模型基准测试

Hongli Zhou, Hui Huang, Ziqing Zhao, Lvyuan Han, Huicheng Wang, Kehai Chen, Muyun Yang, Wei Bao, Jian Dong, Bing Xu, Conghui Zhu, Hailong Cao, Tiejun Zhao

AI总结 本文提出PSN-IRT框架，用于评估大语言模型基准的有效性，揭示现有基准的测量缺陷，并展示如何构建更符合人类偏好的小型基准。

Comments Accepted to AAAI 2026 (Oral)