Research
Cinnamonlab conducts cutting-edge research in computational linguistics, interpretability, and social aspects of AI using a diverse range of interdisciplinary methods and perspectives. We publish in ACL conferences and workshops as well as other top-tier venues in computational linguistics, theoretical linguistics, and AI.
Research Themes
Analysis of Representations
probing, feature attribution, causality
Analyzing what information is encoded by neural representations and how they drive model behavior
Linguistic Evaluation
behavioral tests, psycholinguistic modeling
Assessing the linguistic abilities of language models and comparing them to human language processing
Theory of Architectures
formal language theory, computational complexity
Describing the expressive power of architectures for neural networks and linguistic theories
Theoretical Linguistics
syntax, phonology, psycholinguistics
Gaining insights about human language and linguistics using computational techniques
AI and Society
bias, fairness, humanities and social science
Understanding AI’s relation to society; applying AI to the humanities and social sciences
Foundations of CL/NLP
position papers, metatheory, philosophy of science
Reflecting on what computational linguistics is, why we do it, and how we ought to do it
Publications
2026
Context-Free Recognition with Transformers
arXiv
·
Jan 5, 2026
·
https://arxiv.org/abs/2601.01754
2025
ModelCitizens: Representing Community Voices in Online Safety
EMNLP
·
Nov 5, 2025
·
doi:10.18653/v1/2025.emnlp-main.1571
Generative linguistics, Large Language Models, and the social nature of scientific success
Italian Journal of Linguistics
·
Jul 1, 2025
·
[no id info]
What Goes Into a LM Acceptability Judgment? Rethinking the Impact of Frequency and Length
NAACL
·
May 1, 2025
·
doi:10.18653/v1/2025.naacl-long.109
ERAS: Evaluating the Robustness of Chinese NLP Models to Morphological Garden Path Errors
NAACL
·
Apr 30, 2025
·
doi:10.18653/v1/2025.naacl-long.159
2024
Reflecting the Male Gaze: Quantifying Female Objectification in 19th and 20th Century Novels
LREC-Coling
·
May 22, 2024
·
[no id info]
Universal Generation for Optimality Theory Is PSPACE-Complete
CL
·
Mar 1, 2024
·
doi:10.1162/coli_a_00494
2023
Verb Conjugation in Transformers Is Determined by Linear Encodings of Subject Number
EMNLP Findings
·
Dec 8, 2023
·
doi:10.18653/v1/2023.findings-emnlp.300
Beyond the imitation game: Quantifying and extrapolating the capabilities of language models
TMLR
·
May 11, 2023
·
[no id info]
2022
Formal Language Recognition by Hard Attention Transformers: Perspectives from Circuit Complexity
TACL
·
Jul 27, 2022
·
doi:10.1162/tacl_a_00490
Theory and Applications of Attribution for Interpretable Language Technology
PhD Dissertation
·
Jul 1, 2022
·
[no id info]
An Adversarial Benchmark for Fake News Detection Models
AAAI-22 AdvML Workshop
·
Feb 28, 2022
·
arxiv:2201.00912
2020
Evaluating Attribution Methods using White-Box LSTMs
BlackboxNLP
·
Nov 20, 2020
·
doi:10.18653/v1/2020.blackboxnlp-1.28
Probabilistic Predictions of People Perusing: Evaluating Metrics of Language Model Performance for Psycholinguistic Modeling
CMCL
·
Nov 19, 2020
·
doi:10.18653/v1/2020.cmcl-1.10
Rhythmic Syncope in Subregular Phonology
Penn Linguistics Conference
·
Oct 1, 2020
·
[no id info]
Attribution Analysis of Grammatical Dependencies in LSTMs
arXiv
·
May 4, 2020
·
arxiv:2005.00062
Computing Vowel Harmony: The Generative Capacity of Search & Copy
Annual Meeting on Phonology
·
May 2, 2020
·
doi:10.3765/amp.v8i0.4752
Metrical Grids and Generalized Tier Projection
SCiL
·
Jan 4, 2020
·
doi:10.7275/scil.1227
2019
Finite-state Optimality Theory: non-rationality of Harmonic Serialism
Journal of Language Modelling
·
Sep 16, 2019
·
doi:10.15398/jlm.v7i2.210
Action-Sensitive Phonological Dependencies
SIGMORPHON
·
Aug 2, 2019
·
doi:10.18653/v1/W19-4225
Unbounded Stress in Subregular Phonology
SIGMORPHON
·
Aug 2, 2019
·
doi:10.18653/v1/W19-4216
Finding Hierarchical Structure in Neural Stacks Using Unsupervised Parsing
BlackboxNLP
·
Aug 1, 2019
·
doi:10.18653/v1/W19-4823
Learnability and Overgeneration in Computational Syntax
SCiL
·
Jan 5, 2019
·
doi:10.7275/1qmz-bg76
2018
Context-Free Transductions with Neural Stacks
BlackboxNLP
·
Nov 1, 2018
·
doi:10.18653/v1/W18-5433
2017
Harmonic Serialism and Finite-State Optimality Theory
FSMNLP
·
Jan 1, 2017
·
doi:10.18653/v1/W17-4003
Important: The main result of this paper is incorrect. Please see Hao (2019) for the corrected version.
2015
Model-Theoretic Minimalism
BA Thesis
·
May 14, 2015
·
[no id info]
2014
The peril of sounding manly: A look at vocal characteristics of lawyers before the United States Supreme Court
LabPhon
·
Jul 27, 2014
·
[no id info]
2013
The Dimension and Entropy of ω-Languages
Manuscript
·
Jul 29, 2013
·
[no id info]
Invited Talks
2025
Towards a Linguistics of Large Language Models
Harvard University
·
Oct 21, 2025
Towards a Linguistics of Large Language Models
Stony Brook University
·
Oct 17, 2025
Towards a Science and Mathematics of Language Models
Mathematics of Language Conference
·
Aug 17, 2025
Towards a Linguistics of Large Language Models
Boston University
·
Mar 21, 2025
Towards a Linguistics of Large Language Models
University College London
·
Mar 4, 2025
Towards a Linguistics of Large Language Models
UC San Diego
·
Jan 30, 2025
2024
Word Embeddings: Examining Culture Through a Data-Driven Lens
Vanderbilt University
·
Apr 3, 2024
2023
Verb Conjugation in Transformers Is Determined by Linear Encodings of Subject Number
FLaNN
·
Dec 18, 2023
Transformers and Circuit Complexity
Flatiron Institute
·
Nov 17, 2023
2022
Understanding RNNs and Transformers using Formal Languages
ETH Zürich
·
Nov 21, 2022
Understanding RNNs and Transformers using Formal Languages
University of Notre Dame
·
Nov 7, 2022
Finding Interpretable Word Embedding Subspaces using Covariance and Correlation Maximization
New York University
·
Sep 14, 2022