The Closeness of In-Context Learning and Weight Shifting for Softmax Regression

Neural Information Processing Systems 

Large language models (LLMs) are known for their exceptional performance in natural language processing, making them highly effective in many human life-related tasks.