Contextual Counting: A Mechanistic Study of Transformers on a Quantitative Task
원문 초록
Transformers have revolutionized machine learning across diverse domains, yet understanding their behavior remains crucial, particularly in high-stakes applications. This paper introduces the contextual counting task, a novel toy problem aimed at enhancing our understanding of Transformers in quantitative and scientific contexts. This task requires precise localization and computation within datasets, akin to object detection or region-based scientific analysis. We present theoretical and empirical analysis using both causal and non-causal Transformer architectures, investigating the influence of various positional encodings on performance and interpretability. In particular, we find that causal attention is much better suited for the task, and that no positional embeddings lead to the best accuracy, though rotary embeddings are competitive and easier to train. We also show that out of distribution performance is tightly linked to which tokens it uses as a bias term.
함께 살펴볼 논문
공통 태그를 근거로 연결했습니다. 실제 인용 관계나 논문의 우수성을 뜻하지 않습니다.
Beyond One Shot, Beyond One Perspective: Cross-View and Long-Horizon Distillation for Better LiDAR Representations
MambaFusion: Height-Fidelity Dense Global Fusion for Multi-modal 3D Object Detection
Weakly-supervised Contrastive Learning with Quantity Prompts for Moving Infrared Small Target Detection
Improve Underwater Object Detection through YOLOv12 Architecture and Physics-informed Augmentation
Seg-R1: Segmentation Can Be Surprisingly Simple with Reinforcement Learning
LASFNet: A Lightweight Attention-Guided Self-Modulation Feature Fusion Network for Multimodal Object Detection
토론
이 논문과 연결된 포럼 글이 아직 없습니다.