Experimental Contexts Can Facilitate Robust Semantic Property Inference in Language Models, but Inconsistently

TL;DR

Experimental contexts like in-context examples and instructions can improve language models' ability to infer semantic properties, but this improvement is inconsistent and sometimes relies on superficial heuristics rather than true understanding.

Contribution

This study systematically examines how experimental contexts influence language models' semantic property inference, revealing both potential and limitations.

Findings

Contexts can induce property inheritance in LMs

Inconsistencies arise with different reformulations

Some LMs rely on shallow heuristics rather than semantic understanding

Abstract

Recent zero-shot evaluations have highlighted important limitations in the abilities of language models (LMs) to perform meaning extraction. However, it is now well known that LMs can demonstrate radical improvements in the presence of experimental contexts such as in-context examples and instructions. How well does this translate to previously studied meaning-sensitive tasks? We present a case-study on the extent to which experimental contexts can improve LMs' robustness in performing property inheritance -- predicting semantic properties of novel concepts, a task that they have been previously shown to fail on. Upon carefully controlling the nature of the in-context examples and the instructions, our work reveals that they can indeed lead to non-trivial property inheritance behavior in LMs. However, this ability is inconsistent: with a minimal reformulation of the task, some LMs were found to pick up on shallow, non-semantic heuristics from their inputs, suggesting that the computational principles of semantic property inference are yet to be mastered by LMs.