Can Language Models Compose Skills In-Context?
This addresses the problem of in-context learning for composite tasks in language models, which is incremental as it builds on existing probing methods.
The paper investigates whether language models can compose basic skills from in-context examples to perform composite tasks, finding that models often struggle with this, leading to negative impacts on performance, and proposes a method that improves results by aligning examples with composition steps.
Composing basic skills from simple tasks to accomplish composite tasks is crucial for modern intelligent systems. We investigate the in-context composition ability of language models to perform composite tasks that combine basic skills demonstrated in in-context examples. This is more challenging than the standard setting, where skills and their composition can be learned in training. We conduct systematic experiments on various representative open-source language models, utilizing linguistic and logical tasks designed to probe composition abilities. The results reveal that simple task examples can have a surprising negative impact on the performance, because the models generally struggle to recognize and assemble the skills correctly, even with Chain-of-Thought examples. Theoretical analysis further shows that it is crucial to align examples with the corresponding steps in the composition. This inspires a method for the probing tasks, whose improved performance provides positive support for our insights.