Towards an Analysis of Discourse and Interactional Pragmatic Reasoning Capabilities of Large Language Models

We give a nonexhaustive Within the landscape of linguistic capabilities that overview of the phenomena of those two have been studied and analyzed in Large Language subdomains and the methods traditionally used to Models (LLMs), a considerable amount of research analyze them. We subsequently consider the resulting has focused on phenomena on the level of morphology heterogeneous set of phenomena and methods and syntax (Marvin and Linzen, 2018; as a starting point for our survey of work on discourse Hu et al., 2020). Here, the community seems to pragmatics and interactional pragmatics in have agreed on benchmarks and phenomena that an the context of LLMs. LLM should be capable of (e.g.