# Large Language Models are not Models of Natural Language: they are Corpus Models

arXiv: https://arxiv.org/abs/2112.07055 · HTML: https://mlsanity.com/paper/2112.07055/

## Key facts

- Cited by 6 papers in the map (data through 14 Oct 2026).
- Builds on 17 of its 65 references that are in the map.
- First posted on arXiv 13 Dec 2021.

## Builds on (most cited first, 17 of 17)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding](https://mlsanity.com/paper/1810.04805/) (2018)
- [Language Models are Few-Shot Learners](https://mlsanity.com/paper/2005.14165/) (2020)
- [Evaluating Large Language Models Trained on Code](https://mlsanity.com/paper/2107.03374/) (2021)
- [On the Opportunities and Risks of Foundation Models](https://mlsanity.com/paper/2108.07258/) (2021)
- [Program Synthesis with Large Language Models](https://mlsanity.com/paper/2108.07732/) (2021)
- [The Pile: An 800GB Dataset of Diverse Text for Language Modeling](https://mlsanity.com/paper/2101.00027/) (2020)
- [BERT Rediscovers the Classical NLP Pipeline](https://mlsanity.com/paper/1905.05950/) (2019)
- [Measuring Coding Challenge Competence With APPS](https://mlsanity.com/paper/2105.09938/) (2021)
- [A Primer in BERTology: What we know about how BERT works](https://mlsanity.com/paper/2002.12327/) (2020)
- [Deduplicating Training Data Makes Language Models Better](https://mlsanity.com/paper/2107.06499/) (2021)
- [CCNet: Extracting High Quality Monolingual Datasets from Web Crawl Data](https://mlsanity.com/paper/1911.00359/) (2019)
- [Are Transformers universal approximators of sequence-to-sequence functions?](https://mlsanity.com/paper/1912.10077/) (2019)
- [Learning Word Vectors for 157 Languages](https://mlsanity.com/paper/1802.06893/) (2018)
- [Assessing BERT's Syntactic Abilities](https://mlsanity.com/paper/1901.05287/) (2019)
- [Learning to Execute Programs with Instruction Pointer Attention Graph Neural Networks](https://mlsanity.com/paper/2010.12621/) (2020)
- [Correlating neural and symbolic representations of language](https://mlsanity.com/paper/1905.06401/) (2019)

## Built on by (most cited first, 6 of 6)

- [Position: Key Claims in LLM Research Have a Long Tail of Footnotes](https://mlsanity.com/paper/2308.07120/) (2023)
- [A Sentence is Worth a Thousand Pictures: Can Large Language Models Understand Hum4n L4ngu4ge and the W0rld behind W0rds?](https://mlsanity.com/paper/2308.00109/) (2023)
- [A Review of the Challenges with Massive Web-mined Corpora Used in Large Language Models Pre-Training](https://mlsanity.com/paper/2407.07630/) (2024)
- [Evaluating Large Language Models on the GMAT: Implications for the Future of Business Education](https://mlsanity.com/paper/2401.02985/) (2024)
- [The Linguistic Blind Spot of Value-Aligned Agency, Natural and Artificial](https://mlsanity.com/paper/2207.00868/) (2022)
- [The Quo Vadis of the Relationship between Language and Large Language Models](https://mlsanity.com/paper/2310.11146/) (2023)

## Abstract (arXiv)

Natural Language Processing (NLP) has become one of the leading application areas in the current Artificial Intelligence boom. Transfer learning has enabled large deep learning neural networks trained on the language modeling task to vastly improve performance in almost all downstream language tasks. Interestingly, when the language models are trained with data that includes software code, they demonstrate remarkable abilities in generating functioning computer code from natural language specifications. We argue that this creates a conundrum for the claim that eliminative neural models are a radical restructuring in our understanding of cognition in that they eliminate the need for symbolic abstractions like generative phrase structure grammars. Because the syntax of programming languages is by design determined by phrase structure grammars, neural models that produce syntactic code are apparently uninformative about the theoretical foundations of programming languages. The demonstration that neural models perform well on tasks that involve clearly symbolic systems, proves that they cannot be used as an argument that language and other cognitive systems are not symbolic. Finally, we argue as a corollary that the term language model is misleading and propose the adoption of the working term corpus model instead, which better reflects the genesis and contents of the model.