Everything Everywhere All at Once: LLMs can In-Context Learn Multiple Tasks in Superposition
Year: 2024
Type: preprint
Abstract: Large Language Models (LLMs) have demonstrated remarkable in-context learning (ICL) capabilities. In this study, we explore a surprising phenomenon related to ICL: LLMs can perform multiple, computationally distinct ICL tasks simultaneously, during a single inference call, a capability we term "task superposition". We provide empirical evidence of this phenomenon across various LLM families and scales and show that this phenomenon emerges even if we train the model to in-context learn one task at a time. We offer theoretical explanations that this capability is well within the expressive power of transformers. We also explore how LLMs internally compose task vectors during superposition. Furthermore, we show that larger models can solve more ICL tasks in parallel, and better calibrate their output distribution. Our findings offer insights into the latent capabilities of LLMs, further substantiate the perspective of "LLMs as superposition of simulators", and raise questions about the mechanisms enabling simultaneous task execution.
Source: arXiv (Cornell University)
Authors: Xiong, Zheyang, Cai, Ziyang, Cooper, John, Ge, Albert, Papageorgiou, Vasilis, Sifakis, Zack, Giannou, Angeliki, Lin, Ziqian, Yang, Liu, Agarwal, Saurabh, Chrysos, Grigorios G, Oymak, Samet, Lee, Kangwook, Papailiopoulos, Dimitris
Language: English
Topic: Topic Modeling
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W4403346599