NeurIPS

AMVICC: A Novel Benchmark for Cross-Modal Failure Mode Profiling for VLMs and IGMs

AMVICC: 一种用于VLM和IGM跨模态故障模式分析的新型基准

Aahana Basappa, Pranay Goel, Anusri Karra, Anish Karra, Asa Gilmore, Kevin Zhu

机构 * Centennial High School, Frisco, Texas, USA（Centennial High School, Texas, USA） ； Lebanon Trail High School, Frisco, Texas, USA（Lebanon Trail High School, Texas, USA） ； West Windsor-Plainsboro High School, Princeton Junction, New Jersey, USA（West Windsor-Plainsboro High School, New Jersey, USA） ； Algoverse AI Research, Palo Alto, California, USA（Algoververse AI Research, California, USA） 纠错

AI总结 提出AMVICC基准，通过图像到文本和文本到图像任务系统比较多模态大模型和图像生成模型的视觉推理失败模式，发现故障模式在模型和模态间共享，但存在特定于模型和模态的失败。

Comments 14 pages, 4 figures, 8 tables. Presented at the 39th Conference on Neural Information Processing Systems Workshop: VLM4RWD. Presented at the 43th International Conference on Machine Learning Workshops: ICML 2026 CTB, ICML 2026 FAGEN, ICML 2026 EMM-QA. Authors Aahana Basappa and Pranay Goel contributed equally. Code: https://github.com/AahanaB24/AMVICC, Data: https://doi.org/10.5281/zenodo.17646068

GeoRanker: Distance-Aware Ranking for Worldwide Image Geolocalization

GeoRanker：面向全球图像地理定位的距离感知排序

Pengyue Jia, Seongheon Park, Song Gao, Xiangyu Zhao, Sharon Li

机构 * Department of Data Science, City University of Hong Kong（城市大学数据科学系） ； Department of Computer Sciences, University of Wisconsin-Madison（威斯康星大学麦迪逊分校计算机科学系） ； Department of Geography, University of Wisconsin-Madison（威斯康星大学麦迪逊分校地理系） 纠错

AI总结 提出GeoRanker框架，利用大视觉语言模型联合编码查询-候选交互并预测地理邻近性，引入多阶距离损失以建模结构化空间关系，在IM2GPS3K和YFCC4K基准上达到最优。

Comments NeurIPS 2025

Why Pool When You Can Flow? Active Learning with GFlowNets

为何池选，不如流选？基于GFlowNets的主动学习

Renfei Zhang, Mohit Pandey, Artem Cherkasov, Martin Ester

机构 * School of Computer Science, Simon Fraser University, Burnaby, BC, Canada（Simon Fraser大学计算机科学学院，Burnaby, BC, Canada） ； Vancouver Prostate Centre, University of British Columbia, Vancouver, BC, Canada（温哥华前列腺中心，不列颠哥伦比亚大学，Vancouver, BC, Canada） ； Faculty of Medicine, University of British Columbia, Vancouver, BC, Canada（不列颠哥伦比亚大学医学院，Vancouver, BC, Canada） ； Diagen AI 纠错

AI总结 提出BALD-GFlowNet框架，用生成流网络直接采样高信息分子，替代传统池选，实现与池大小无关的可扩展性，在虚拟筛选中达到与BALD相当的性能并生成更多样化分子。

Comments Accepted at the NeurIPS 2025 Workshop on AI Virtual Cells and Instruments: A New Era in Drug Discovery and Development (AI4D3 2025), San Diego, California, USA. 6 pages; 5 figures