mlsanityData to 14 Oct 2026

On the Similarity of Circuits across Languages: a Case Study on the Subject-verb Agreement Task

arXiv 2410.06496 ↗First posted 9 Oct 2024cs.CLcited by 24

Not on the map. Why?

Abstract (arXiv)

Several algorithms implemented by language models have recently been successfully reversed-engineered. However, these findings have been concentrated on specific tasks and models, leaving it unclear how universal circuits are across different settings. In this paper, we study the circuits implemented by Gemma 2B for solving the subject-verb agreement task across two different languages, English and Spanish. We discover that both circuits are highly consistent, being mainly driven by a particular attention head writing a `subject number' signal to the last residual stream, which is read by a small set of neurons in the final MLPs. Notably, this subject number signal is represented as a direction in the residual stream space, and is language-independent. We demonstrate that this direction has a causal effect on the model predictions, effectively flipping the Spanish predicted verb number by intervening with the direction found in English. Finally, we present evidence of similar behavior in other models within the Gemma 1 and Gemma 2 families.

Key facts. Cited by 24 papers in the map (data through 14 Oct 2026). Builds on 20 of its 38 references that are in the map. First posted on arXiv 9 Oct 2024.

Builds on

20 of 38 references.

References · 20

  1. 2018BERT: Pre-training of Deep Bidirectional Transformers for Language Understandingcited by 40,399
  2. 2020Language Models are Few-Shot Learnerscited by 32,419
  3. 2022PaLM: Scaling Language Modeling with Pathwayscited by 5,326
  4. 2022Locating and Editing Factual Associations in GPTcited by 2,691
  5. 2023Inference-Time Intervention: Eliciting Truthful Answers from a Language Modelcited by 1,302
  6. 2022Interpretability in the Wild: a Circuit for Indirect Object Identification in GPT-2 smallcited by 1,165
  7. 2021Probing Classifiers: Promises, Shortcomings, and Advancescited by 876
  8. 2023Dissecting Recall of Factual Associations in Auto-Regressive Language Modelscited by 528
  9. 2021Causal Abstractions of Neural Networkscited by 422
  10. 2016Assessing the Ability of LSTMs to Learn Syntax-Sensitive Dependenciescited by 354
  11. 2024Do Llamas Work in English? On the Latent Language of Multilingual Transformerscited by 281
  12. 2019Assessing BERT's Syntactic Abilitiescited by 192
8 more
  1. 2023Linear Representations of Sentiment in Large Language Modelscited by 163
  2. 2021Causal Analysis of Syntactic Agreement Mechanisms in Neural Language Modelscited by 131
  3. 2020Finding Universal Grammatical Relations in Multilingual BERTcited by 110
  4. 2019The emergence of number and syntax units in LSTM language modelscited by 94
  5. 2022Interpreting Language Models with Contrastive Explanationscited by 92
  6. 2022Probing for the Usage of Grammatical Numbercited by 59
  7. 2024CausalGym: Benchmarking causal interpretability methods on linguistic taskscited by 46
  8. 2021Sparse Interventions in Language Models with Differentiable Maskingcited by 34

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2024: 1; 2025: 11; 2026: 12.

24 papers.

Sort:

By area:

  1. 2025MultiBLiMP 1.0: A Massively Multilingual Benchmark of Linguistic Minimal Pairscited by 71
  2. 2025Large Language Models Share Representations of Latent Grammatical Concepts Across Typologically Diverse Languagescited by 59
  3. 2025Paths Not Taken: Understanding and Mending the Multilingual Factual Recall Pipelinecited by 21
  4. 2025LLMs as a synthesis between symbolic and distributed approaches to languagecited by 13
  5. 2025Different types of syntactic agreement recruit the same units within large language modelscited by 8
  6. 2025Are Knowledge and Reference in Multilingual Language Models Cross-Lingually Consistent?cited by 5
  7. 2024Thank You, Stingray: Multilingual Large Language Models Can Not (Yet) Disambiguate Cross-Lingual Word Sensecited by 4
  8. 2025Discursive Circuits: How Do Language Models Understand Discourse Relations?cited by 4
  9. 2025Exploring the Translation Mechanism of Large Language Modelscited by 3
  10. 2025Steering Language Models in Multi-Token Generation: A Case Study on Tense and Aspectcited by 2
  11. 2025How Instruction-Tuning Imparts Length Control: A Cross-Lingual Mechanistic Analysiscited by 1
  12. 2026Understanding or Memorizing? A Case Study of German Definite Articles in Language Modelscited by 1
12 more
  1. 2026On the Emergence and Test-Time Use of Structural Information in Large Language Modelscited by 1
  2. 2026The Grammar of Transformers: A Systematic Review of Interpretability Research on Syntactic Knowledge in Language Modelscited by 1
  3. 2026Beyond Language: Format-Agnostic Reasoning Subspaces in Large Language Modelscited by 1
  4. 2026Judge Circuits Explain Format-Induced Inconsistency in LLM-as-a-Judgecited by 1
  5. 2026Test, then Route: How Language Models Execute In-Context Conditional Rules Across Models and Languagescited by 1
  6. 2025Quantification and object perception in Multimodal Large Language Models and human linguistic cognitioncited by 0
  7. 2026A Study of Crosslinguistic Influence in Language Modelscited by 0
  8. 2026Exploring Language-Agnosticity in Function Vectors: A Case Study in Machine Translationcited by 0
  9. 2026Shared Circuits for Shared Grammar: Tracing Subject-Verb Agreement Across Languagescited by 0
  10. 2026Causal Interventions Reveal Typologically Organized Syntactic Mechanisms in Multilingual Language Modelscited by 0
  11. 2026Shared circuits predict whether LLMs generalize across formats in arithmetic reasoningcited by 0
  12. 2026FTB Graph: Determining and Validating First-token Broadcasters and Language-Identity Head Circuits in Multilingual Language Modelscited by 0