Revisiting Compositional Generalization Capability of Large Language Models Considering Instruction Following Ability

TL;DR

This paper introduces Ordered CommonGen, a benchmark to evaluate large language models' ability to follow instructions and generalize compositionally, revealing current models' limitations in ordered concept generation.

Contribution

The paper proposes a new benchmark for assessing instruction-following and compositional generalization in LLMs, providing a comprehensive analysis of 36 models' performance.

Findings

LLMs often produce low-diversity outputs biased toward specific concept orders.

Even the best models achieve only about 75% ordered coverage.

Current models need improvements in instruction-following and compositional generalization.

Abstract

In generative commonsense reasoning tasks such as CommonGen, generative large language models (LLMs) compose sentences that include all given concepts. However, when focusing on instruction-following capabilities, if a prompt specifies a concept order, LLMs must generate sentences that adhere to the specified order. To address this, we propose Ordered CommonGen, a benchmark designed to evaluate the compositional generalization and instruction-following abilities of LLMs. This benchmark measures ordered coverage to assess whether concepts are generated in the specified order, enabling a simultaneous evaluation of both abilities. We conducted a comprehensive analysis using 36 LLMs and found that, while LLMs generally understand the intent of instructions, biases toward specific concept order patterns often lead to low-diversity outputs or identical results even when the concept order is altered. Moreover, even the most instruction-compliant LLM achieved only about 75% ordered coverage, highlighting the need for improvements in both instruction-following and compositional generalization capabilities.