The Stochastic Parrot on LLM's Shoulder: A Summative Assessment of Physical Concept Understanding
This research addresses the problem of LLMs' lack of true understanding for users relying on these models for accurate information, particularly in areas requiring physical concept comprehension.
This study investigated the physical concept understanding of large language models (LLMs), finding that state-of-the-art models lag behind humans by approximately 40%. The results also demonstrated the presence of the 'stochastic parrot' phenomenon, where LLMs fail to understand concepts despite being able to describe and recognize them in natural language.
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.