Can In-context Learners Learn a Reasoning Concept from Demonstrations?

Warning

This publication doesn't include Faculty of Economics and Administration. It includes Faculty of Informatics. Official publication website can be found on muni.cz.
Authors

ŠTEFÁNIK Michal KADLČÍK Marek

Year of publication 2023
Type Article in Proceedings
Conference Proceedings of the 1st Workshop on Natural Language Reasoning and Structured Explanations (NLRSE)
MU Faculty or unit

Faculty of Informatics

Citation
Keywords in-context learning; few-shot learning; generalization
Description Language models exhibit an emergent ability to learn a new task from a small number of input-output demonstrations. However, recent work shows that in-context learners largely rely on their pre-trained knowledge, such as the sentiment of the labels, instead of learning new associations from the input. We argue that the commonly-used few-shot evaluation using a random selection of in-context demonstrations can not disentangle models' reliance on such biases, as most of the randomly-selected demonstrations do not present relations informative for prediction beyond exposing the task's input-output distribution. Therefore, to evaluate models' in-context learning ability independent of models' memory, we introduce a Concept-sharing few-shot learning method choosing the demonstrations that share an underlying concept with the predicted sample. We extract a set of such concepts from available human explanations and measure how much models can benefit from presenting these concepts in few-shot demonstrations. We find that most of the recent in-context learners can not consistently benefit from the demonstrated concepts, irrespective of the model size. However, we note that T0 models are more sensitive to exhibited concepts, benefiting from concept-sharing demonstrations in 7 out of 8 evaluation scenarios.
Related projects:

You are running an old browser version. We recommend updating your browser to its latest version.