AAAI

SkipCat: Rank-Maximized Low-Rank Compression of Large Language Models via Shared Projection and Block Skipping

SkipCat: 通过共享投影和块跳过实现大型语言模型的秩最大化低秩压缩

Yu-Chen Lu, Sheng-Feng Yu, Hui-Hsien Weng, Pei-Shuo Wang, Yu-Fang Hu, Liang Hung-Chun, Hung-Yueh Chiang, Kai-Chiang Wu

AI总结 SkipCat通过共享投影和块跳过技术实现低秩压缩，无需额外微调即可在相同压缩率下提升零样本任务的准确率7%。

Comments Accepted by AAAI 2026

Towards a Common Framework for Autoformalization

迈向自动形式化的一个共同框架

Agnieszka Mensfelt, David Tena Cucala, Santiago Franco, Angeliki Koutsoukou-Argyraki, Vince Trencsenyi, Kostas Stathis

机构 * Department of Computer Science, Royal Holloway, University of London（伦敦大学皇家霍洛威学院计算机科学系） ； Department of Computer Science and Technology, University of Cambridge（剑桥大学计算机科学与技术系） 纠错

AI总结 本文旨在提出一个统一框架，以促进不同领域之间的交叉融合，推动下一代人工智能系统的发展。

Comments Presented at NeLaMKRR@KR, 2025 (arXiv:2511.09575). A shorter version of this work will appear in the Proceedings of the AAAI Conference on Artificial Intelligence (AAAI 2026)

AI Copilots for Reproducibility in Science: A Case Study

人工智能辅助科学可重复性：案例研究

Adrien Bibal, Steven N. Minton, Deborah Khider, Yolanda Gil

AI总结 本文提出了一种AI驱动的可重复性助手，通过生成Jupyter笔记本和建议，帮助科学家更高效地实现计算重现，减少重现时间并提高结果的透明度。

Comments Reproducible Artificial Intelligence (RAI2026) Workshop, AAAI 2026

Less Is More: Sparse and Cooperative Perturbation for Point Cloud Attacks

少即是多：用于点云攻击的稀疏与协作扰动

Keke Tang, Tianyu Hao, Xiaofei Wang, Weilong Peng, Denghui Zhang, Peican Zhu, Zhihong Tian

AI总结 本文提出SCP框架，通过稀疏且协作的点扰动方法，实现高攻击成功率和更强的对抗效果，适用于点云攻击场景。

Comments Accepted by AAAI'2026 (Oral)

Towards Test-time Efficient Visual Place Recognition via Asymmetric Query Processing

通过非对称查询处理实现测试时高效的视觉位置识别

Jaeyoon Kim, Yoonki Cho, Sung-Eui Yoon

AI总结 本文提出一种非对称VPR框架，通过地理记忆库和隐式嵌入增强技术，实现低计算成本的高效视觉位置识别。

Comments AAAI 2026

JoDiffusion: Jointly Diffusing Image with Pixel-Level Annotations for Semantic Segmentation Promotion

JoDiffusion: 通过像素级标注联合扩散图像以促进语义分割

Haoyu Wang, Lei Zhang, Wenrui Liu, Dengyang Jiang, Wei Wei, Chen Ding

AI总结 JoDiffusion通过联合扩散图像与像素级标注，提升语义分割的性能和可扩展性。

Comments Accepted at AAAI 2026

Unified Interactive Multimodal Moment Retrieval via Cascaded Embedding-Reranking and Temporal-Aware Score Fusion

通过级联嵌入-重排序和时间感知得分融合实现统一的交互式多模态时刻检索

Toan Le Ngo Thanh, Phat Ha Huu, Tan Nguyen Dang Duy, Thong Nguyen Le Minh, Anh Nguyen Nhu Tinh

AI总结 本文提出通过级联嵌入-重排序和时间感知得分融合实现统一的多模态时刻检索系统，解决跨模态噪声、时间序列连贯性和模态选择问题。

Comments Accepted at AAAI Workshop 2026

StegaVAR: Privacy-Preserving Video Action Recognition via Steganographic Domain Analysis

StegaVAR：通过隐写领域分析实现隐私保护的视频动作识别

Lixin Chen, Chaomeng Chen, Jiale Zhou, Zhijian Wu, Xun Lin

AI总结 StegaVAR通过隐写领域分析实现视频动作识别的隐私保护，提出STeP和CroDA方法提升隐写分析效果。

Comments 13 pages, 10 figures. This is the extended version of the paper accepted at AAAI 2026, including related works and appendix

ViInfographicVQA: A Benchmark for Single and Multi-image Visual Question Answering on Vietnamese Infographics

ViInfographicVQA：单图和多图越南信息图视觉问答基准测试

Tue-Thu Van-Dinh, Hoang-Duy Tran, Truong-Binh Duong, Mai-Hanh Pham, Binh-Nam Le-Nguyen, Quoc-Thai Nguyen

机构 * Neu.edu.vn 纠错

AI总结 ViInfographicVQA是首个针对越南信息图VQA的基准测试，通过单图和多图任务评估模型在复杂信息图中的视觉问答能力。

Comments 10 pages, 4 figures, Accepted to AI4Research @ AAAI

Task-Specific Distance Correlation Matching for Few-Shot Action Recognition

面向少样本动作识别的任务特定距离相关匹配

Fei Long, Yao Zhang, Jiaming Lv, Jiangtao Xie, Peihua Li

AI总结 TS-FSAR通过任务特定距离相关匹配和改进的CLIP微调框架，提升少样本动作识别的性能。

Comments 9 pages. 4 figures, conference;Accepted by AAAI 2026

From Imitation to Discrimination: Toward A Generalized Curriculum Advantage Mechanism Enhancing Cross-Domain Reasoning Tasks

从模仿到辨别：一种通用的课程优势机制，增强跨领域推理任务

Changpeng Yang, Jinyang Wu, Yuchen Liu, Shuai Zhang, Yang Li, Qiliang Liang, Hongzhen Wang, Shuai Nie, Jiaming Xu, Runyu Shi, Ying Huang, Guoquan Zhang

机构 * \equalcontrib（1号机构） 纠错

AI总结 CAPO是一种基于优势信号的自适应课程机制，通过引导模仿学习和引入负信号提升跨领域推理任务的泛化能力。

Comments Accepted by AAAI 2026

EfficientFlow: Efficient Equivariant Flow Policy Learning for Embodied AI

EfficientFlow: 为具身AI高效的等价流策略学习

Jianlei Chang, Ruofeng Mei, Wei Ke, Xiangyu Xu

机构 * Xi'an Jiaotong University（西安交通大学） 纠错

AI总结 EfficientFlow通过引入等价性与加速正则化策略，提升具身AI的效率与性能。

Comments Accepted by AAAI 2026. Project Page: https://efficientflow.github.io/

Rethinking Long-tailed Dataset Distillation: A Uni-Level Framework with Unbiased Recovery and Relabeling

重新思考长尾数据集蒸馏：一种统一层面框架，具有无偏恢复和重标记

Xiao Cui, Yulei Qin, Xinyue Li, Wengang Zhou, Hongsheng Li, Houqiang Li

AI总结 本文提出一种统一框架，通过无偏恢复和重标记方法提升长尾数据集蒸馏效果，实验显示在多个基准上显著提高准确率。

Comments AAAI 2026 (Oral)

Coherent Multi-Agent Trajectory Forecasting in Team Sports with CausalTraj

在团队体育中利用因果轨迹进行相干多智能体轨迹预测

Wei Zhen Teoh

AI总结 CausalTraj通过因果轨迹模型在团队体育中实现多智能体轨迹预测，提升联合预测的准确性和场景连贯性。

Comments 9 pages, 3 figures, accepted to the AI4TS Workshop at AAAI 2026

CoordAR: One-Reference 6D Pose Estimation of Novel Objects via Autoregressive Coordinate Map Generation

CoordAR：通过自回归坐标图生成实现新型物体的单参考6D姿态估计

Dexin Zuo, Ang Li, Wei Wang, Wenxian Yu, Danping Zou

AI总结 CoordAR通过自回归坐标图生成方法，实现对新型物体的单参考6D姿态估计，提升了对称性和遮挡等挑战的鲁棒性。

Comments 7 pages, accepted by AAAI 2026 (oral)

SproutBench: A Benchmark for Safe and Ethical Large Language Models for Youth

SproutBench: 为青少年设计的安全和伦理大型语言模型基准

Wenpeng Xing, Lanyi Wei, Haixiao Hu, Jingyi Yu, Rongchang Li, Mohan Li, Changting Lin, Meng Han

AI总结 SproutBench通过1283个发展性对抗提示评估47种LLMs，揭示儿童安全漏洞，为青少年AI设计提供指导。

Comments Accepted in AAAI 2026 Workshop on AI for Education

Incoherence as Oracle-less Measure of Error in LLM-Based Code Generation

不一致性作为无 oracle 的错误度量方法在基于 LLM 的代码生成中的应用

Thomas Valentin, Ardi Madadi, Gaetano Sapia, Marcel Böhme

AI总结 本文提出了一种无 oracle 的错误度量方法，用于评估基于LLM的代码生成的正确性，通过不一致性度量有效识别错误程序并替代传统 oracle 评估。

Comments Accepted at AAAI'26 (extended version). 8 pages + refs and appendix

Journal ref 40th Annual AAAI Conference on Artificial Intelligence (AAAI), 2026

MR-COSMO: Visual-Text Memory Recall and Direct CrOSs-MOdal Alignment Method for Query-Driven 3D Segmentation

MR-COSMO：一种用于查询驱动3D分割的视觉-文本记忆召回与直接跨模态对齐方法

Chade Li, Pengju Zhang, Yihong Wu

AI总结 MR-COSMO通过视觉-文本记忆召回与直接跨模态对齐方法，在查询驱动的3D分割中实现几何与语义特征的精确融合，提升点云分割性能。

Comments Accepted by AAAI 2026. Copyright (c) 2026, Association for the Advancement of Artificial Intelligence (www.aaai.org). All rights reserved

Multimodal DeepResearcher: Generating Text-Chart Interleaved Reports From Scratch with Agentic Framework

多模态深度研究员：从零开始生成文本-图表交织报告的代理框架

Zhaorui Yang, Bo Pan, Han Wang, Yiyao Wang, Xingyu Liu, Luoxuan Weng, Yingchaojie Feng, Haozhe Feng, Minfeng Zhu, Bo Zhang, Wei Chen

AI总结 多模态深度研究员通过代理框架实现从零开始生成文本-图表交织报告，利用FDV结构化文本表示提升可视化生成质量。

Comments AAAI 2026 Oral

A Flat Minima Perspective on Understanding Augmentations and Model Robustness

从平坦极小值角度理解增强和模型鲁棒性

Weebum Yoo, Sung Whan Yoon

AI总结 本文从平坦极小值角度理论揭示了标签保持增强提升模型鲁棒性的通用条件，并通过实验验证其在不同分布偏移下的有效性。

Comments In Proceedings of the Association for the Advancement of Artificial Intelligence (AAAI) 2026, Singapore

Relation-R1: Progressively Cognitive Chain-of-Thought Guided Reinforcement Learning for Unified Relation Comprehension

Relation-R1: 逐步认知链式思维引导的强化学习用于统一关系理解

Lin Li, Wei Chen, Jiahui Li, Kwang-Ting Cheng, Long Chen

AI总结 Relation-R1通过整合认知链式思维引导的强化学习，实现了对二元和N元关系的统一理解，提升了视觉-语义定位能力。

Comments AAAI 2026

PIMRL: Physics-Informed Multi-Scale Recurrent Learning for Burst-Sampled Spatiotemporal Dynamics

PIMRL：基于突发采样的多尺度递归学习用于时空动态

Han Wan, Qi Wang, Yuan Mi, Rui Zhang, Hao Sun

AI总结 PIMRL通过物理引导的多尺度递归学习，有效处理突发采样的稀疏时空动态数据，显著提升建模精度和效率。

Comments To appear in AAAI 2026

An Overall Real-Time Mechanism for Classification and Quality Evaluation of Rice

一种综合的实时机制用于水稻分类和质量评估

Wanke Xia, Ruoxin Peng, Haoqi Chu, Xinlei Zhu, Zhiyu Yang, Yiting Zhao, Lili Yang

机构 * Wanke Xia 1,2（夏万克 1,2） ； Ruoxin Peng 1（彭鲁新 1） ； Haoqi Chu 1（褚浩琪 1） ； Xinlei Zhu 1（朱新磊 1） ； Zhiyu Yang 1（杨志宇 1） ； Yiting Zhao 2（赵毅婷 2） ； Lili Yang 1（杨丽丽 1） 纠错

AI总结 本研究提出一种结合目标检测、深度神经网络和传统机器学习的实时机制，用于高效准确地评估水稻品种、完整度和糊粉度。

Comments Accepted at AAAI 2026 Workshop (AgriAI)

DABL: Detecting Semantic Anomalies in Business Processes Using Large Language Models

基于大语言模型的业务流程语义异常检测：DABL

Wei Guan, Jian Cao, Jianqi Gao, Haiyan Zhao, Shiyou Qian

AI总结 DABL利用大语言模型检测业务流程中的语义异常，通过生成正常轨迹和模拟异常，提升泛化能力和解释性。

Journal ref Proceedings of the AAAI Conference on Artificial Intelligence (AAAI), vol. 39, no. 11, pp. 11735-11744, 2025

A Benchmark Dataset for Spatially Aligned Road Damage Assessment in Small Uncrewed Aerial Systems Disaster Imagery

用于小无人 aerial 系统灾害影像中空间对齐道路损坏评估的基准数据集

Thomas Manzini, Priyankari Perali, Raisa Karnik, Robin R. Murphy

AI总结 本文提出一个用于小无人 aerial 系统灾害影像中道路损坏评估的基准数据集，并通过空间对齐技术提升模型性能。

Comments 11 pages, 6 figures, 6 tables. To appear AAAI'26

The Forecast Critic: Leveraging Large Language Models for Poor Forecast Identification

预测批评者：利用大语言模型进行差预测识别

Luke Bhan, Hanyu Zhang, Andrew Gordon Wilson, Michael W. Mahoney, Chuck Arvin

机构 * Amazon（亚马逊） 纠错

AI总结 利用大语言模型进行预测监控，通过识别不合理预测提升预测质量。

Comments Presented at AAAI 2026 AI4TS workshop and AABA4ET workshop

Structured Personalization: Modeling Constraints as Matroids for Data-Minimal LLM Agents

结构化个性化：将约束建模为Matroid以实现数据最小的LLM代理

Daniel Platnick, Marjan Alirezaie, Hossein Rahnama

AI总结 本文提出通过Matroid建模约束，实现数据最小的LLM个性化方法，解决现实中的结构化约束问题。

Comments Accepted to the AAAI 2026 Workshop on Personalization in the Era of Large Foundation Models (PerFM), 5 pages, 1 figure

PANDA -- Patch And Distribution-Aware Augmentation for Long-Tailed Exemplar-Free Continual Learning

PANDA -- 无示例持续学习中的补丁与分布感知增强

Siddeshwar Raghavan, Jiangpeng He, Fengqing Zhu

AI总结 PANDA通过补丁与分布感知增强方法，解决无示例持续学习中的任务间不平衡问题，提升学习效果和泛化能力。

Comments Accepted in AAAI 2026 Main Technical Track

WDT-MD: Wavelet Diffusion Transformers for Microaneurysm Detection in Fundus Images

WDT-MD：用于视网膜图像微动脉瘤检测的小波扩散变换器

Yifei Sun, Yuzhi He, Junhao Jia, Jinhong Wang, Ruiquan Ge, Changmiao Wang, Hongxia Xu

AI总结 WDT-MD 通过小波扩散变换器框架，解决微动脉瘤检测中的身份映射、区分困难和重建效果差问题，提升视网膜图像筛查性能。

Comments 9 pages, 6 figures, 8 tables, accepted by AAAI 2026

Debiasing Machine Learning Predictions for Causal Inference Without Additional Ground Truth Data: "One Map, Many Trials" in Satellite-Driven Poverty Analysis

消除机器学习预测偏差以进行因果推断而无需额外真实数据：卫星驱动的贫困分析中的'一张地图，多次试验'

Markus B. Pettersson, Connor T. Jerzak, Adel Daoud

AI总结 本文提出两种无需额外数据的后处理方法，用于减少机器学习预测中的收缩偏见，从而提升因果推断的准确性，适用于贫困分析等多领域应用。

Comments To appear in the Proceedings of AAAI 2026