Provable Limitations of Acquiring Meaning from Ungrounded Form: What Will Future Language Models Understand?
Open accessThe DOI matches its Crossref record; 26 of the 27 references it cites were found in a registry or an index.
Abstract
Language models trained on billions of tokens have recently led to unprecedented results on many NLP tasks. This success raises the question of whether, in principle, a system can ever “understand” raw text without access to some form of grounding. We formally investigate the abilities of ungrounded systems to acquire meaning. Our analysis focuses on the role of “assertions”: textual contexts that provide indirect clues about the underlying semantics. We study whether assertions enable a system to emulate representations preserving semantic relations like equivalence. We find that assertions enable semantic emulation of languages that satisfy a strong notion of semantic transparency. However, for classes of languages where the same expression can take different values in different contexts, we show that emulation can become uncomputable. Finally, we discuss differences between our formal model and natural language, exploring how our results generalize to a modal setting and other semantic relations. Together, our results suggest that assertions in code or language do not provide sufficient signal to fully emulate semantic representations. We formalize ways in which ungrounded language models appear to be fundamentally limited in their ability to “understand”.
Overview
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“[85] propose a proof that no isomorphism can exist between an ungrounded LM and a formal semantic representation of meaning.”
From Symbols and grounding in large language models (2023) · cites this paper as [85]
Experimental — accuracy not yet measured
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
Devlin, J., Chang, M.-W., Lee, K., & Toutanova, K. (2019) · Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers) · doi:10.18653/v1/n19-1423 · Cited by 54,325 · cited 1×
“Recently, language models trained on huge datasets of raw text have pushed the limits of natural language processing (Devlin et al., 2019; Raffel et al., 2019; Brown et al., 2020, among others).”
BackgroundLanguage Models are Few-Shot Learners
Brown, T. B., Mann, B., Ryder, N., Subbiah, M., Kaplan, J., Dhariwal, P., Neelakantan, A., Shyam, P., Sastry, G., Askell, A., et al. (2020) · arXiv · doi:10.48550/arxiv.2005.14165 · Cited by 23,372
Grice, H. P. (1975) · Speech Acts · doi:10.1163/9789004368811_003 · Cited by 23,221 · cited 1×
“Indeed, this is one of Grice’s Maxims (Grice, 1975): a set of basic principles proposed to govern the pragmatics of natu-ral language.”
BackgroundExploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer
Raffel, C., Shazeer, N., Roberts, A., Lee, K., Narang, S., Matena, M., Zhou, Y., Li, W., & Liu, P. J. (2019) · arXiv · doi:10.48550/arxiv.1910.10683 · Cited by 10,929 · cited 1×
“Recently, language models trained on huge datasets of raw text have pushed the limits of natural language processing (Devlin et al., 2019; Raffel et al., 2019; Brown et al., 2020, among others).”
BackgroundI.—COMPUTING MACHINERY AND INTELLIGENCE
TURING, A. M. (1950) · Mind · doi:10.1093/mind/lix.236.433 · Cited by 10,290 · cited 1×
“Turing (1950) constitutes an early behaviorist perspective; more recent approaches tend to emphasize not just an ex-ternal view of a system’s behavior, but also “how it is achieved” (Levesque, 2014).”
BackgroundOn computable numbers, with an application to the entscheidungsproblem
Turing, A. M. (1937) · Proceedings of the London Mathematical Society · doi:10.1112/plms/s2-42.1.230 · Cited by 8,283
Computing Machinery and Intelligence (1950)
Turing, A. (2004) · The Essential Turing · doi:10.1093/oso/9780198250791.003.0017 · Cited by 6,118
Harris, Z. S. (1954) · WORD · doi:10.1080/00437956.1954.11659520 · Cited by 4,807
GLUE: A Multi-Task Benchmark and Analysis Platform for Natural Language Understanding
Wang, A., Singh, A., Michael, J., Hill, F., Levy, O., & Bowman, S. (2018) · Proceedings of the 2018 EMNLP Workshop BlackboxNLP: Analyzing and Interpreting Neural Networks for NLP · doi:10.18653/v1/w18-5446 · Cited by 4,222 · cited 1×
“Understanding can be behaviorally diagnosed in neural models by evaluating them on benchmarks (Wang et al., 2018).”
BackgroundLearning regular sets from queries and counterexamples
Angluin, D. (1987) · Information and Computation · doi:10.1016/0890-5401(87)90052-6 · Cited by 2,114
Semantics in generative grammar
Heim, I., & Kratzer, A. (1998) · Cited by 1,905
A primer in BERTology: What we know about how BERT works
Rogers, A., Kovaleva, O., & Rumshisky, A. (2020) · Transactions of the Association for Computational Linguistics · doi:10.1162/tacl_a_00349 · Cited by 1,552 · cited 1×
“Rather, analysis of massive language models has revealed that, to some degree, knowledge of syntactic and semantic dependencies can emerge without explicit supervision (Rogers et al., 2020; Tenney et al., 2019).”
BackgroundBERT Rediscovers the Classical NLP Pipeline
Tenney, I., Das, D., & Pavlick, E. (2019) · Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics · doi:10.18653/v1/p19-1452 · Cited by 1,307 · cited 1×
“Rather, analysis of massive language models has revealed that, to some degree, knowledge of syntactic and semantic dependencies can emerge without explicit supervision (Rogers et al., 2020; Tenney et al., 2019).”
BackgroundClimbing towards NLU: On Meaning, Form, and Understanding in the Age of Data
Bender, E. M., & Koller, A. (2020) · Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics · doi:10.18653/v1/2020.acl-main.463 · Cited by 1,082 · cited 12×
“In this work, we formalized an argument that was raised as Bender and Koller (2020) as a thought experiment. Bender and Koller (2020) question whether unsupervised training objectives are the right goal to chase for achieving natural language understanding.”
BackgroundMethodInfluentialWhat you can cram into a single $&!#* vector: Probing sentence embeddings for linguistic properties
Conneau, A., Kruszewski, G., Lample, G., Barrault, L., & Baroni, M. (2018) · Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · doi:10.18653/v1/p18-1198 · Cited by 657 · cited 1×
“An alternate approach is probing (Adi et al., 2017; Conneau et al., 2018; Hupkes and Zuidema, 2018; Hewitt and Liang, 2019; Belinkov and Glass, 2019), which investigates how directly a model’s representations encode semantic relations by measuring if they can be easily decoded from them.”
BackgroundAnalysis Methods in Neural Language Processing: A Survey
Belinkov, Y., & Glass, J. (2019) · Transactions of the Association for Computational Linguistics · doi:10.1162/tacl_a_00254 · Cited by 477 · cited 1×
“An alternate approach is probing (Adi et al., 2017; Conneau et al., 2018; Hupkes and Zuidema, 2018; Hewitt and Liang, 2019; Belinkov and Glass, 2019), which investigates how directly a model’s representations encode semantic relations by measuring if they can be easily decoded from them.”
BackgroundDesigning and Interpreting Probes with Control Tasks
Hewitt, J., & Liang, P. (2019) · Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP) · doi:10.18653/v1/d19-1275 · Cited by 393 · cited 1×
“An alternate approach is probing (Adi et al., 2017; Conneau et al., 2018; Hupkes and Zuidema, 2018; Hewitt and Liang, 2019; Belinkov and Glass, 2019), which investigates how directly a model’s representations encode semantic relations by measuring if they can be easily decoded from them.”
BackgroundFine-grained Analysis of Sentence Embeddings Using Auxiliary Prediction Tasks
Adi, Y., Kermany, E., Belinkov, Y., Lavi, O., & Goldberg, Y. (2016) · arXiv (Cornell University) · doi:10.48550/arxiv.1608.04207 · Cited by 283 · cited 2×
“An alternate approach is probing (Adi et al., 2017; Conneau et al., 2018; Hupkes and Zuidema, 2018; Hewitt and Liang, 2019; Belinkov and Glass 2019), which investigates how directly a model’s representations encode semantic relations by measuring if they can be easily decoded from Figure 1: An illustration of Definiti…”
BackgroundSelf-citationFine-grained Analysis of Sentence Embeddings Using Auxiliary Prediction Tasks
Adi, Y. (2024) · arXiv (Cornell University) · doi:10.57702/tcr06vve · Cited by 212
Levesque, H. J. (2014) · Artificial Intelligence · doi:10.1016/j.artint.2014.03.007 · Cited by 105 · cited 1×
“Turing (1950) constitutes an early behaviorist perspective; more recent approaches tend to emphasize not just an ex-ternal view of a system’s behavior, but also “how it is achieved” (Levesque, 2014).”
BackgroundWinoWhy: A Deep Diagnosis of Essential Commonsense Knowledge for Answering Winograd Schema Challenge
Zhang, H., Zhao, X., & Song, Y. (2020) · Proceedings of the conference - Association for Computational Linguistics. Meeting · doi:10.18653/v1/2020.acl-main.508 · Cited by 71
Three Learnable Models for the Description of Language
Clark, A. (2010) · Lecture notes in computer science · doi:10.1007/978-3-642-13089-2_2 · Cited by 49 · cited 2×
“Resembling Clark (2010), we refer to a tuple 〈l, r〉 ∈ (Σ∗)2 as a syntactic context.”
BackgroundHupkes, D., & Zuidema, W. H. (2018) · International Joint Conference on Artificial Intelligence · doi:10.24963/ijcai.2018/796 · Cited by 46 · cited 1×
“An alternate approach is probing (Adi et al., 2017; Conneau et al., 2018; Hupkes and Zuidema, 2018; Hewitt and Liang, 2019; Belinkov and Glass, 2019), which investigates how directly a model’s representations encode semantic relations by measuring if they can be easily decoded from them.”
BackgroundPratt-Hartmann, I. E., & Third, A. (2006) · Notre Dame Journal of Formal Logic · doi:10.1305/ndjfl/1153858644 · Cited by 27
Winter, Y. (2016) · Edinburgh University Press eBooks · doi:10.1515/9780748677771 · Cited by 27
Winter, Y. (2016) · Cited by 14
Research works printed but not linked: 6
- Chomsky, 1965, Aspects of the Theory of Syntax
- Horn, 2020, Negation
- Pratt-Hartmann, 2006, More fragments of language, Notre Dame Journal of Formal Logic, 47, 151, 10.1305/ndjfl/1153858644 · DOI
- Fintel, 2011, Intensional semantics, Unpublished Lecture Notes
- Whitehead, 1925–1927, Principia Mathematica
- Yogatama, 2019, Learning and evaluating general linguistic intelligence. arXiv: 1901 .11373
Cited by
On the Opportunities and Risks of Foundation Models
Bommasani, R., Hudson, D. A., Adeli, E., Altman, R., Arora, S., von Arx, S., Bernstein, M. S., Bohg, J., Bosselut, A., Brunskill, E., et al. (2021) · arXiv · doi:10.48550/arxiv.2108.07258 · Cited by 2,676
Text Data Augmentation for Deep Learning
Shorten, C., Khoshgoftaar, T. M., & Furht, B. (2021) · Journal of Big Data · doi:10.1186/s40537-021-00492-0 · Cited by 1,740
Symbolic Knowledge Distillation: from General Language Models to Commonsense Models
West, P., Bhagavatula, C., Hessel, J., Hwang, J., Jiang, L., Le Bras, R., Lu, X., Welleck, S., & Choi, Y. (2022) · Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies · doi:10.18653/v1/2022.naacl-main.341 · Cited by 156
“Prior works have suggested that pre-trained language models possess limited understanding of commonsense knowledge (Merrill et al., 2021; Talmor et al., 2021; Davis and Marcus, 2017) despite (1)We will share this following the anonymity period.”
BackgroundSymbols and grounding in large language models
Pavlick, E. (2023) · Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences · doi:10.1098/rsta.2022.0041 · Cited by 95
“[85] propose a proof that no isomorphism can exist between an ungrounded LM and a formal semantic representation of meaning.”
BackgroundDoes ChatGPT have semantic understanding? A problem with the statistics-of-occurrence strategy
Titus, L. M. (2024) · Cognitive Systems Research · doi:10.1016/j.cogsys.2023.101174 · Cited by 82
Metadata
- Authors
- Merrill, W., Goldberg, Y., Schwartz, R., & Smith, N. A.
- Affiliations
- Allen Institute for Artificial Intelligence (US)
- Bar-Ilan University (IL)
- University of Washington (US)
- Venue
- Transactions of the Association for Computational Linguistics
- Publisher
- MIT Press
- Published
- 2021
- Volume
- 9
- Pages
- 1047-1060
- ISSN
- 2307-387X (ISSN-L)
- OpenAlex
- W3156509887
- Semantic Scholar
- 2bb1e1a5b9a16f6828fe94736cea5dab264533a6
- Cited by
- 44
- References
- 27
- Citations by year
- 2026: 6 · 2025: 3 · 2024: 1 · 2023: 14 · 2022: 16 · 2021: 4
- FWCI
- 34.56
- Type
- journal article
- Language
- en
- Open access
- diamond
- doi.org
- Locations
- Transactions of the Association for Computational Linguistics (publishedVersion, cc-by)
- arXiv (Cornell University) (submittedVersion)
- DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
- arXiv (Cornell University)
- Funders
- Allen Institute for Artificial Intelligence
- Topics
- Social Sciences › Arts and Humanities › Language and Linguistics › Syntax, Semantics, Linguistic Variation
- Physical Sciences › Computer Science › Artificial Intelligence › Logic, Reasoning, and Knowledge
- Physical Sciences › Computer Science › Artificial Intelligence › Natural Language Processing Techniques
- Keywords
- meaning acquisition, assertion, semantic transparency, uncomputability, semantic relations, equivalence, modal logic, formal semantics, language comprehension
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.