MOS: Towards Scaling Out-of-distribution Detection for Large Semantic Space

2021-05-05 · CVPR 2021 1 · Rui Huang, Yixuan Li

Detecting out-of-distribution (OOD) inputs is a central challenge for safely deploying machine learning models in the real world. Existing solutions are mainly driven by small datasets, with low resolution and very few class labels (e.g., CIFAR). As a result, OOD detection for large-scale image classification tasks remains largely unexplored. In this paper, we bridge this critical gap by proposing a group-based OOD detection framework, along with a novel OOD scoring function termed MOS. Our key idea is to decompose the large semantic space into smaller groups with similar concepts, which allows simplifying the decision boundaries between in- vs. out-of-distribution data for effective OOD detection. Our method scales substantially better for high-dimensional class space than previous approaches. We evaluate models trained on ImageNet against four carefully curated OOD datasets, spanning diverse semantics. MOS establishes state-of-the-art performance, reducing the average FPR95 by 14.33% while achieving 6x speedup in inference compared to the previous best method.

📄 PDF Abstract BibTeX arXiv:2105.01879

Code (5)

deeplearning-wisc/large_scale_ood 공식 구현 pytorch

deeplearning-wisc/gradnorm_ood pytorch

ma-kjh/CMA-OoDD pytorch

tmlr-group/class_prior pytorch

tmlr-group/neglabel pytorch

Tasks

image-classification Image Classification Out-of-Distribution Detection

Evaluating the Impact of Model Scale for Compositional Generalization in Semantic Parsing

2022-05-24 · Linlu Qiu, Peter Shaw, Panupong Pasupat, Tianze Shi 외

Despite their strong performance on many tasks, pre-trained language models have been shown to struggle on out-of-distribution compositional generalization. Meanwhile, recent work has shown considerable improvements on m…

Decoder In-Context Learning Language Modelling Semantic Parsing+1

Zero-Shot Event Detection by Multimodal Distributional Semantic Embedding of Videos

2015-12-02 · Mohamed Elhoseiny, Jingen Liu, Hui Cheng, Harpreet Sawhney 외

We propose a new zero-shot Event Detection method by Multi-modal Distributional Semantic embedding of videos. Our model embeds object and action concepts as well as other available modalities from videos into a distribut…

Event Detection

OstQuant: Refining Large Language Model Quantization with Orthogonal and Scaling Transformations for Better Distribution Fitting

2025-01-23 · Xing Hu, Yuan Cheng, Dawei Yang, Zukang Xu 외

Post-training quantization (PTQ) has emerged as a widely adopted technique for compressing and accelerating Large Language Models (LLMs). The major challenge in LLM quantization is that uneven and heavy-tailed data distr…

Language Modeling Language Modelling Large Language Model Quantization

Beyond the Training Distribution: Mapping Generalization Boundaries in Neural Program Synthesis

2026-04-30 · Henrik Voigt, Michael Habeck, Joachim Giesen arxiv

Large-scale transformers achieve impressive results on program synthesis benchmarks, yet their true generalization capabilities remain obscured by data contamination and opaque training corpora. To rigorously assess whet…

Program Synthesis

Latent Semantic Manifolds in Large Language Models

2026-03-17 · Mohamed A. Mabrok arxiv

Large Language Models (LLMs) perform internal computations in continuous vector spaces yet produce discrete tokens -- a fundamental mismatch whose geometric consequences remain poorly understood. We develop a mathematica…

Model Compression