The Grammar of Transformers: A Systematic Review of Interpretability Research on Syntactic Knowledge in Language Models
cs.CL, cs.AI
Submitted: 2026-01-09
Updated: 2026-08-31
Comments: Published as a main conference paper at EMNLP 2026
License: http://creativecommons.org/licenses/by/4.0/
The gist: We present a systematic review of 337 articles evaluating the syntactic abilities of Transformer-based language models (TLMs), reporting on over 3,000 datapoints spanning a wide range of syntactic
Terminology
Abstract
We present a systematic review of 337 articles evaluating the syntactic abilities of Transformer-based language models (TLMs), reporting on over 3,000 datapoints spanning a wide range of syntactic phenomena, languages, models, and methods. We take the data to collectively show that TLMs encode a non-trivial amount of syntactic knowledge. Behavioral evidence shows strong performance on formal syntactic phenomena, but weaker and more variable performance on phenomena at the syntax-semantics interface. Performance is also consistently lower for languages with less digital support. Probing and mechanistic studies further support the presence of syntactic knowledge in TLMs. Yet, because most work remains observational and methodologically heterogeneous, insight into the detailed computational mechanisms underlying syntactic processing remains limited. At the same time, the literature remains heavily concentrated on English and BERT-like models. We discuss the implications of our results and provide recommendations for future research.
Sources
- Interpreting token compositionality in LLMs: A robustness analysis
- Differential syntactic and semantic encoding in LLMs
- Does BERT agree? Evaluating knowledge of structure dependence through agreement relations
- Constituency Parsing using LLMs
- LLMs as a synthesis between symbolic and distributed approaches to language
- Language Model Behavior: A Comprehensive Survey
- When does word order matter and when doesn't it?
- How Syntax Specialization Emerges in Language Models
- Tree Transformers are an Ineffective Model of Syntactic Constituency
- Deep Clustering of Text Representations for Supervision-free Probing of Syntax
- Large Language Models as Neurolinguistic Subjects: Discrepancy between Performance and Competence
- Do Attention Heads in BERT Track Syntactic Dependencies?
- MultiBLiMP 1.0: A Massively Multilingual Benchmark of Linguistic Minimal Pairs
- Manipulating language models' training data to study syntactic constraint learning: the case of English passivization
- Incremental Comprehension of Garden-Path Sentences by Large Language Models: Semantic Interpretation, Syntactic Re-Analysis, and Attention
- Syntax Representation in Word Embeddings and Neural Networks -- A Survey
- ChatGPT is a Potential Zero-Shot Dependency Parser
- Linguistic Interpretability of Transformer-based Language Models: a systematic review
- Language Models as Models of Language
- minicons: Enabling Flexible Behavioral and Representational Analyses of Transformer Language Models
Related papers
- Exploring Solution Divergence and Its Effect on Large Language Model Problem Solving
- Ishigaki-IDS-Bench: A Benchmark for Generating Information Delivery Specification from BIM Information Requirements
- Subliminal Steering: Stronger Encoding of Hidden Signals
- MedStruct-S: A Benchmark for Key Discovery, Key-Conditioned QA and Semi-Structured Extraction from OCR Clinical Reports
- The End of Transformers? On Challenging Attention and the Rise of Sub-Quadratic Architectures
- Untangling the Mechanisms of Misleading Context in Medical Question Answering