Do Syntax Heads Influence Model Behavior?
Not on the map. Why?
Abstract (arXiv)
Transformer language models develop attention heads that specialize in syntactic relations, for example, attending from verbs to their subjects. Do these heads play a causal role in the model's syntactic predictions? Evidence for such a role is mixed to negative: even when we force these heads to attend to a distractor noun other than the subject, the model's syntactic predictions are still more consistent with the subject than with the distractor. Interventions on specialized syntax heads had a significantly stronger effect on syntactic predictions than interventions on random heads, but only in one of the two measures we propose.
Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 13 of its 23 references that are in the map. First posted on arXiv 13 Oct 2026.
Builds on
13 of 23 references.
References · 13
- 2017Attention Is All You Needcited by 60,361
- 2023Pythia: A Suite for Analyzing Large Language Models Across Training and Scalingcited by 1,786
- 2019What Does BERT Look At? An Analysis of BERT's Attentioncited by 1,003
- 2019Analyzing Multi-Head Self-Attention: Specialized Heads Do the Heavy Lifting, the Rest Can Be Prunedcited by 856
- 2016Assessing the Ability of LSTMs to Learn Syntax-Sensitive Dependenciescited by 354
- 2018Colorless green recurrent networks dream hierarchicallycited by 213
- 2019Assessing BERT's Syntactic Abilitiescited by 192
- 2020A Systematic Assessment of Syntactic Generalization in Neural Language Modelscited by 165
- 2020Probing the Probing Paradigm: Does Probing Accuracy Entail Task Relevance?cited by 151
- 2021Causal Analysis of Syntactic Agreement Mechanisms in Neural Language Modelscited by 131
- 2023Sudden Drops in the Loss: Syntax Acquisition, Phase Transitions, and Simplicity Bias in MLMscited by 124
- 2022Probing for the Usage of Grammatical Numbercited by 59
Built on by
Not cited yet.