What If the Input is Expanded in OOD Detection?

Authors: Boxuan Zhang, Jianing Zhu, Zengmao Wang, Tongliang Liu, Bo Du, Bo Han

NeurIPS 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments and analyses have been conducted to understand and verify the effectiveness of Co Ver.
Researcher Affiliation Academia 1School of Computer Science, Wuhan University 2TMLR Group, Department of Computer Science, Hong Kong Baptist University 3Sydney AI Center, The University of Sydney 4RIKEN Center for Advanced Intelligence Project
Pseudocode No The paper does not include any explicitly labeled pseudocode or algorithm blocks.
Open Source Code Yes The code is publicly available at: https://github.com/tmlr-group/Co Ver.
Open Datasets Yes Following previous work [1, 31], we adopt the Image Net-1K OOD benchmark [24], which uses the Image Net-1K [14] as ID data and i Naturalist [49], SUN [55], Places [60], and Textures [7] as OOD data.
Dataset Splits Yes To select the most effective corruption types for each method, we use SVHN [37] as the validation set.
Hardware Specification Yes All experiments are conducted on NVIDIA Ge Force RTX 3090 GPUs with Python 3.10 and Py Torch 2.2.
Software Dependencies Yes All experiments are conducted on NVIDIA Ge Force RTX 3090 GPUs with Python 3.10 and Py Torch 2.2.
Experiment Setup Yes By default, we use the Co Ver score in the max-softmax form and set τ = 1 as the temperature.