StructTest: Benchmarking LLMs' Reasoning through Compositional Structured Outputs

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Chen, Hailin, Jiao, Fangkai, Ravaut, Mathieu, Farruque, Nawshad, Nguyen, Xuan Phi, Qin, Chengwei, Dey, Manan, Ding, Bosheng, Xiong, Caiming, Joty, Shafiq, Zhou, Yingbo
Formato: Preprint
Publicado: 2024
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866917962708942848
author Chen, Hailin
Jiao, Fangkai
Ravaut, Mathieu
Farruque, Nawshad
Nguyen, Xuan Phi
Qin, Chengwei
Dey, Manan
Ding, Bosheng
Xiong, Caiming
Joty, Shafiq
Zhou, Yingbo
author_facet Chen, Hailin
Jiao, Fangkai
Ravaut, Mathieu
Farruque, Nawshad
Nguyen, Xuan Phi
Qin, Chengwei
Dey, Manan
Ding, Bosheng
Xiong, Caiming
Joty, Shafiq
Zhou, Yingbo
contents The rapid advancement of large language models (LLMs) demands robust, unbiased, and scalable evaluation methods. However, human annotations are costly to scale, model-based evaluations are susceptible to stylistic biases, and target-answer-based benchmarks are vulnerable to data contamination and cheating. To address these limitations, we propose StructTest, a novel benchmark that evaluates LLMs on their ability to follow compositional instructions and generate structured outputs, providing an unbiased, cost-effective, and difficult-to-cheat evaluation framework. Assessments are conducted deterministically using a rule-based evaluator, which can be easily extended to new tasks and datasets. By testing structured outputs across diverse domains including Summarization, Code, HTML, and Math, and evaluating 17 popular LLMs, we demonstrate that StructTest remains challenging even for top-performing models like Deepseek-V3/R1 and GPT-4o, establishing it as a robust proxy for measuring reasoning capabilities. We believe StructTest offers a critical and complementary approach to achieving objective and comprehensive model evaluation.
format Preprint
id arxiv_https___arxiv_org_abs_2412_18011
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle StructTest: Benchmarking LLMs' Reasoning through Compositional Structured Outputs
Chen, Hailin
Jiao, Fangkai
Ravaut, Mathieu
Farruque, Nawshad
Nguyen, Xuan Phi
Qin, Chengwei
Dey, Manan
Ding, Bosheng
Xiong, Caiming
Joty, Shafiq
Zhou, Yingbo
Computation and Language
The rapid advancement of large language models (LLMs) demands robust, unbiased, and scalable evaluation methods. However, human annotations are costly to scale, model-based evaluations are susceptible to stylistic biases, and target-answer-based benchmarks are vulnerable to data contamination and cheating. To address these limitations, we propose StructTest, a novel benchmark that evaluates LLMs on their ability to follow compositional instructions and generate structured outputs, providing an unbiased, cost-effective, and difficult-to-cheat evaluation framework. Assessments are conducted deterministically using a rule-based evaluator, which can be easily extended to new tasks and datasets. By testing structured outputs across diverse domains including Summarization, Code, HTML, and Math, and evaluating 17 popular LLMs, we demonstrate that StructTest remains challenging even for top-performing models like Deepseek-V3/R1 and GPT-4o, establishing it as a robust proxy for measuring reasoning capabilities. We believe StructTest offers a critical and complementary approach to achieving objective and comprehensive model evaluation.
title StructTest: Benchmarking LLMs' Reasoning through Compositional Structured Outputs
topic Computation and Language
url https://arxiv.org/abs/2412.18011