SpeciaLex: A Benchmark for In-Context Specialized Lexicon Learning

TL;DR

SpeciaLex is a comprehensive benchmark designed to evaluate large language models' ability to understand and apply specialized lexicons across various tasks, aiding the development of more precise content generation tools.

Contribution

Introduces SpeciaLex, a new benchmark with 18 tasks and 1,785 instances, to assess LLMs' handling of specialized lexicon constraints in content tasks.

Findings

Larger models generally perform better on the benchmark.

Open-source models show competitive performance with closed-source counterparts.

Model recency and setup significantly influence performance.

Abstract

Specialized lexicons are collections of words with associated constraints such as special definitions, specific roles, and intended target audiences. These constraints are necessary for content generation and documentation tasks (e.g., writing technical manuals or children's reading materials), where the goal is to reduce the ambiguity of text content and increase its overall readability for a specific group of audience. Understanding how large language models can capture these constraints can help researchers build better, more impactful tools for wider use beyond the NLP community. Towards this end, we introduce SpeciaLex, a benchmark for evaluating a language model's ability to follow specialized lexicon-based constraints across 18 diverse subtasks with 1,785 test instances covering core tasks of Checking, Identification, Rewriting, and Open Generation. We present an empirical evaluation of 15 open and closed-source LLMs and discuss insights on how factors such as model scale, openness, setup, and recency affect performance upon evaluating with the benchmark.