Evaluating Large Language Models on Controlled Generation Tasks

TL;DR

This paper analyzes the controllability of large language models in generation tasks, revealing they often struggle with fine-grained constraints compared to smaller, fine-tuned models.

Contribution

It provides an extensive benchmark analysis of large language models' ability to meet fine-grained constraints in controlled generation tasks.

Findings

Large language models often fall behind smaller models in fine-grained control.

Models are comparable or exceed smaller models in some benchmarks.

Struggle to meet hard, fine-grained constraints.

Abstract

While recent studies have looked into the abilities of large language models in various benchmark tasks, including question generation, reading comprehension, multilingual and etc, there have been few studies looking into the controllability of large language models on generation tasks. We present an extensive analysis of various benchmarks including a sentence planning benchmark with different granularities. After comparing large language models against state-of-the-start finetuned smaller models, we present a spectrum showing large language models falling behind, are comparable, or exceed the ability of smaller models. We conclude that **large language models struggle at meeting fine-grained hard constraints**.