To understand neural network behavior, recent works quantitatively compare different networks' learned representations using canonical correlation analysis (CCA), centered kernel alignment (CKA), and other dissimilarity measures. Unfortunately, these widely used measures often disagree on fundamental observations, such as whether deep networks differing only in random initialization learn similar representations. These disagreements raise the question: which, if any, of these dissimilarity measures should we believe? We provide a framework to ground this question through a concrete test: measures should have sensitivity to changes that affect functional behavior, and specificity against changes that do not. We quantify this through a variety of functional behaviors including probing accuracy and robustness to distribution shift, and examine changes such as varying random initialization and deleting principal components. We find that current metrics exhibit different weaknesses, note that a classical baseline performs surprisingly well, and highlight settings where all metrics appear to fail, thus providing a challenge set for further improvement.
Does Location Matter? A Black-Box Causal Audit Reveals Model-Dependent Audio Grounding in LALM Hallucination
Sharif University of Technology(谢里夫理工大学);Amirkabir University of Technology(阿米尔卡比尔理工大学);New Uzbekistan University(新乌兹别克斯坦大学);Okinawa Institute of Science and Technology(冲绳科学技术大学院大学)
Semantically Guided Adversarial Testing of Vision Models Using Language Models
Institute of Theoretical and Applied Informatics, Polish Academy of Sciences(波兰科学院理论与应用信息学研究所);University of Lille, CNRS, Inria, Centrale Lille, UMR 9189 CRIStAL(里尔大学)
Quantum Similarity Testing with Convolutional Neural Networks
The University of Hong Kong(香港大学);National University of Singapore(新加坡国立大学);Zhejiang University(浙江大学);Perimeter Institute for Theoretical Physics(圆周理论物理研究所)
Few-Shot Scenario Testing for Autonomous Vehicles Based on Neighborhood Coverage and Similarity
PKU-WUHAN Institute for Artificial Intelligence(北京大学-武汉人工智能研究院);Tsinghua University(清华大学);Beijing National Research Center for Information Science and Technology (BNRist)(北京信息科学与技术国家研究中心);Tsinghua-Berkeley Shenzhen Institute (TBSI)(清华-伯克利深圳学院);Hebei Expressway Group Limited(河北高速公路集团有限公司)
Predicting Machine Translation Performance on Low-Resource Languages: The Role of Domain Similarity
University of Michigan(密歇根大学);University of Toronto(多伦多大学);Centro de Investigación en Matemáticas(数学研究中心);Amherst College(阿默斯特学院);Universiteit Gent(根特大学);Ontario Tech University(安大略理工大学)
ReSi: A Comprehensive Benchmark for Representational Similarity Measures
University of Passau(帕绍大学);German Cancer Research Center (DKFZ)(德国癌症研究中心);Helmholtz Imaging(亥姆霍兹成像中心);University of Heidelberg(海德堡大学);University of Mannheim(曼海姆大学);RWTH Aachen University(亚琛工业大学);Heidelberg University Hospital(海德堡大学医院);National Center for Tumor Diseases (NCT) Heidelberg(海德堡国家肿瘤疾病中心);GESIS - Leibniz Institute for the Social Sciences(莱布尼茨社会科学研究所GESIS);Complexity Science Hub(复杂性科学中心)