Towards Compositionally Generalizable Semantic Parsing in Large Language Models: A Survey

TL;DR

This survey addresses the challenge of compositional generalization in semantic parsing for large language models, focusing on unseen, nested outputs across tasks like task-oriented dialogue and text-to-SQL. It reviews definitions and benchmarks (SCAN, COGS, CFQ) and surveys methodology, analyzes factors limiting generalization (structural biases, decoding bottlenecks, pretraining distribution, and model usage), and categorizes improvement strategies into data augmentation, neuro-symbolic methods, and prompt-based approaches (including decomposition into steps and into facts). The work highlights that effective progress often requires modular, multi-step processing rather than end-to-end solutions, and discusses orthogonal generalization pathways that can be pursued in parallel. Overall, the paper provides a structured synthesis to guide researchers and practitioners toward robust compositional generalization in semantic parsing with LLMs, detailing concrete methods and evaluation considerations for future work.

Abstract

Compositional generalization is the ability of a model to generalize to complex, previously unseen types of combinations of entities from just having seen the primitives. This type of generalization is particularly relevant to the semantic parsing community for applications such as task-oriented dialogue, text-to-SQL parsing, and information retrieval, as they can harbor infinite complexity. Despite the success of large language models (LLMs) in a wide range of NLP tasks, unlocking perfect compositional generalization still remains one of the few last unsolved frontiers. The past few years has seen a surge of interest in works that explore the limitations of, methods to improve, and evaluation metrics for compositional generalization capabilities of LLMs for semantic parsing tasks. In this work, we present a literature survey geared at synthesizing recent advances in analysis, methods, and evaluation schemes to offer a starting point for both practitioners and researchers in this area.