5 papers
Stronger Normalization-Free Transformers
Mingzhi Chen, Taiming Lu, Jiachen Zhu +2
Although normalization layers have long been viewed as indispensable components of deep learning architectures, the recent introduction of Dynamic Tanh (DyT) has demonstrated that…
Command-V: Pasting LLM Behaviors via Activation Profiles
Barry Wang, Avi Schwarzschild, Alexander Robey +4
Retrofitting large language models (LLMs) with new behaviors typically requires full finetuning or distillation-costly steps that must be repeated for every architecture. In this w…
ConSense: Continually Sensing Human Activity with WiFi via Growing and Picking
Rong Li, Tao Deng, Siwei Feng +2
WiFi-based human activity recognition (HAR) holds significant application potential across various fields. To handle dynamic environments where new activities are continuously intr…
Idiosyncrasies in Large Language Models
Mingjie Sun, Yida Yin, Zhiqiu Xu +2
In this work, we unveil and study idiosyncrasies in Large Language Models (LLMs) -- unique patterns in their outputs that can be used to distinguish the models. To do so, we consid…
Massive Values in Self-Attention Modules are the Key to Contextual Knowledge Understanding
Mingyu Jin, Kai Mei, Wujiang Xu +5
Large language models (LLMs) have achieved remarkable success in contextual knowledge understanding. In this paper, we show that these concentrated massive values consistently emer…