English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Humans possess a remarkable ability to assign novel interpretations tolinguistic expressions, enabling them to learn new words and understandcommunity-specific connotations. However, Large Language Models (LLMs) have aknowledge cutoff and are costly to finetune repeatedly. Therefore, it iscrucial for LLMs to learn novel interpretations in-context. In this paper, wesystematically analyse the ability of LLMs to acquire novel interpretationsusing in-context learning. To facilitate our study, we introduce MAGNIFICo, anevaluation suite implemented within a text-to-SQL semantic parsing frameworkthat incorporates diverse tokens and prompt settings to simulate real-worldcomplexity. Experimental results on MAGNIFICo demonstrate that LLMs exhibit asurprisingly robust capacity for comprehending novel interpretations fromnatural language descriptions as well as from discussions within longconversations. Nevertheless, our findings also highlight the need for furtherimprovements, particularly when interpreting unfamiliar words or when composingmultiple novel interpretations simultaneously in the same example.Additionally, our analysis uncovers the semantic predispositions in LLMs andreveals the impact of recency bias for information presented in long contexts.

MAGNIFICo: Evaluating the In-Context Learning Ability of Large Language Models to Generalize to Novel Interpretations