mlsanityData to 14 Oct 2026

Probing for Incremental Parse States in Autoregressive Language Models

arXiv 2211.09748 ↗First posted 17 Nov 2022cs.CL cs.LGcited by 24

Not on the map. Why?

Abstract (arXiv)

Next-word predictions from autoregressive neural language models show remarkable sensitivity to syntax. This work evaluates the extent to which this behavior arises as a result of a learned ability to maintain implicit representations of incremental syntactic structures. We extend work in syntactic probing to the incremental setting and present several probes for extracting incomplete syntactic structure (operationalized through parse states from a stack-based parser) from autoregressive language models. We find that our probes can be used to predict model preferences on ambiguous sentence prefixes and causally intervene on model representations and steer model behavior. This suggests implicit incremental syntactic inferences underlie next-word predictions in autoregressive neural language models.

Key facts. Cited by 24 papers in the map (data through 14 Oct 2026). Builds on 28 of its 47 references that are in the map. First posted on arXiv 17 Nov 2022.

Builds on

28 of 47 references.

References · 28

  1. 2014Adam: A Method for Stochastic Optimizationcited by 37,240
  2. 2021Probing Classifiers: Promises, Shortcomings, and Advancescited by 876
  3. 2019Plug and Play Language Models: A Simple Approach to Controlled Text Generationcited by 799
  4. 2019Designing and Interpreting Probes with Control Taskscited by 640
  5. 2019What do you learn from context? Probing for sentence structure in contextualized word representationscited by 438
  6. 2018Targeted Syntactic Evaluation of Language Modelscited by 242
  7. 2020Information-Theoretic Probing for Linguistic Structurecited by 207
  8. 2021Implicit Representations of Meaning in Neural Language Modelscited by 180
  9. 2020A Systematic Assessment of Syntactic Generalization in Neural Language Modelscited by 165
  10. 2019Neural Language Models as Psycholinguistic Subjects: Representations of Syntactic Statecited by 138
  11. 2020On the Predictive Power of Neural Language Models for Human Real-Time Comprehension Behaviorcited by 119
  12. 2016Recurrent Neural Network Grammarscited by 109
16 more
  1. 2020Syntactic Structure from Deep Learningcited by 99
  2. 2021Counterfactual Interventions Reveal the Causal Effect of Relative Clause Representations on Agreement Predictioncited by 59
  3. 2018Finding Syntax in Human Encephalography with Beam Searchcited by 41
  4. 2021What if This Modified That? Syntactic Interventions via Counterfactual Embeddingscited by 39
  5. 2020Can neural networks acquire a structural bias from raw linguistic data?cited by 35
  6. 2021Sparse Interventions in Language Models with Differentiable Maskingcited by 34
  7. 2021A Targeted Assessment of Incremental Processing in Neural LanguageModels and Humanscited by 32
  8. 2017Effective Inference for Generative Neural Parsingcited by 25
  9. 2021Structural Guidance for Transformer Language Modelscited by 25
  10. 2019Structural Supervision Improves Learning of Non-Local Grammatical Dependenciescited by 24
  11. 2021The Language Model Understood the Prompt was Ambiguous: Probing Syntactic Uncertainty Through Generationcited by 15
  12. 2021Effective Batching for Recurrent Neural Network Grammarscited by 12
  13. 2022When Does Syntax Mediate Neural Language Model Performance? Evidence from Dropout Probescited by 12
  14. 2021Causal Transformers Perform Below Chance on Recursive Nested Constructions, Unlike Humanscited by 8
  15. 2022Probing for Labeled Dependency Treescited by 8
  16. 2017Arc-swift: A Novel Transition System for Dependency Parsingcited by 2

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2023: 6; 2024: 5; 2025: 5; 2026: 8.

24 papers.

Sort:

By area:

  1. 2023Emergent Linear Representations in World Models of Self-Supervised Sequence Modelscited by 319
  2. 2024A polar coordinate system represents syntax in large language modelscited by 25
  3. 2025We Can't Understand AI Using our Existing Vocabularycited by 25
  4. 2023Backpack Language Modelscited by 15
  5. 2024SyntaxShap: Syntax-aware Explainability Method for Text Generationcited by 11
  6. 2025Mechanisms vs. Outcomes: Probing for Syntax Fails to Explain Performance on Targeted Syntactic Evaluationscited by 7
  7. 2023Are Structural Concepts Universal in Transformer Language Models? Towards Interpretable Cross-Lingual Generalizationcited by 6
  8. 2023Multilingual Nonce Dependency Treebanks: Understanding how Language Models represent and process syntactic structurecited by 5
  9. 2025Probing Syntax in Large Language Models: Successes and Remaining Challengescited by 5
  10. 2024Incremental Sentence Processing Mechanisms in Autoregressive Transformer Language Modelscited by 4
  11. 2023Approximating CKY with Transformerscited by 3
  12. 2025Trick or Neat: Adversarial Ambiguity and Language Model Evaluationcited by 3
12 more
  1. 2023On the Challenges of Fully Incremental Neural Dependency Parsingcited by 2
  2. 2026Polar probe linearly decodes semantic structures from LLMscited by 2
  3. 2024When Only Time Will Tell: Interpreting How Transformers Process Local Ambiguities Through the Lens of Restart-Incrementalitycited by 1
  4. 2026The Grammar of Transformers: A Systematic Review of Interpretability Research on Syntactic Knowledge in Language Modelscited by 1
  5. 2026Implicit Representations of Grammaticality in Language Modelscited by 1
  6. 2026Linear representations of grammaticality in neural language modelscited by 1
  7. 2024Multipath parsing in the braincited by 0
  8. 2025Derivational Probing: Unveiling the Layer-wise Derivation of Syntactic Structures in Neural Language Modelscited by 0
  9. 2026Gated Tree Cross-Attention for Checkpoint-Compatible Syntax Injection in Decoder-Only LLMscited by 0
  10. 2026A Mechanism and Optimization Study on the Impact of Information Density on User-Generated Content Named Entity Recognitioncited by 0
  11. 2026Why are language models less surprised than humans? Testing the Parse Multiplicity Mismatch Hypothesiscited by 0
  12. 2026Representation of syntax in LLMs through the lens of linear distance and similarity-aware entropycited by 0