The Stochastic Parrot on LLM's Shoulder: A Summative Assessment of Physical Concept Understanding

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Yu, Mo, Liu, Lemao, Wu, Junjie, Chung, Tsz Ting, Zhang, Shunchi, Li, Jiangnan, Yeung, Dit-Yan, Zhou, Jie
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866913689242697728
author Yu, Mo
Liu, Lemao
Wu, Junjie
Chung, Tsz Ting
Zhang, Shunchi
Li, Jiangnan
Yeung, Dit-Yan
Zhou, Jie
author_facet Yu, Mo
Liu, Lemao
Wu, Junjie
Chung, Tsz Ting
Zhang, Shunchi
Li, Jiangnan
Yeung, Dit-Yan
Zhou, Jie
contents In a systematic way, we investigate a widely asked question: Do LLMs really understand what they say?, which relates to the more familiar term Stochastic Parrot. To this end, we propose a summative assessment over a carefully designed physical concept understanding task, PhysiCo. Our task alleviates the memorization issue via the usage of grid-format inputs that abstractly describe physical phenomena. The grids represents varying levels of understanding, from the core phenomenon, application examples to analogies to other abstract patterns in the grid world. A comprehensive study on our task demonstrates: (1) state-of-the-art LLMs, including GPT-4o, o1 and Gemini 2.0 flash thinking, lag behind humans by ~40%; (2) the stochastic parrot phenomenon is present in LLMs, as they fail on our grid task but can describe and recognize the same concepts well in natural language; (3) our task challenges the LLMs due to intrinsic difficulties rather than the unfamiliar grid format, as in-context learning and fine-tuning on same formatted data added little to their performance.
format Preprint
id arxiv_https___arxiv_org_abs_2502_08946
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle The Stochastic Parrot on LLM's Shoulder: A Summative Assessment of Physical Concept Understanding
Yu, Mo
Liu, Lemao
Wu, Junjie
Chung, Tsz Ting
Zhang, Shunchi
Li, Jiangnan
Yeung, Dit-Yan
Zhou, Jie
Computation and Language
Artificial Intelligence
Computer Vision and Pattern Recognition
Machine Learning
In a systematic way, we investigate a widely asked question: Do LLMs really understand what they say?, which relates to the more familiar term Stochastic Parrot. To this end, we propose a summative assessment over a carefully designed physical concept understanding task, PhysiCo. Our task alleviates the memorization issue via the usage of grid-format inputs that abstractly describe physical phenomena. The grids represents varying levels of understanding, from the core phenomenon, application examples to analogies to other abstract patterns in the grid world. A comprehensive study on our task demonstrates: (1) state-of-the-art LLMs, including GPT-4o, o1 and Gemini 2.0 flash thinking, lag behind humans by ~40%; (2) the stochastic parrot phenomenon is present in LLMs, as they fail on our grid task but can describe and recognize the same concepts well in natural language; (3) our task challenges the LLMs due to intrinsic difficulties rather than the unfamiliar grid format, as in-context learning and fine-tuning on same formatted data added little to their performance.
title The Stochastic Parrot on LLM's Shoulder: A Summative Assessment of Physical Concept Understanding
topic Computation and Language
Artificial Intelligence
Computer Vision and Pattern Recognition
Machine Learning
url https://arxiv.org/abs/2502.08946