publications.bib

@inproceedings{hublog,
      title={How to visualize training dynamics in neural networks}, 
      author={Michael Y. Hu and Shreyans Jain and Sangam Chaulagain and Naomi Saphra},
      year={2025},
	booktitle = {{Blog Post Track at International Conference on Learning Representations (ICLR BlogPosts)}},
}

@inproceedings{
johnson-yu2024understanding,
title={Understanding biological active sensing behaviors by interpreting learned artificial agent policies},
author={Sonja Johnson-Yu and Satpreet Harcharan Singh and Federico Pedraja and Denis Turcu and Pratyusha Sharma and Saphra, Naomi and Nathaniel Sawtell and Kanaka Rajan},
booktitle={Workshop on Interpretable Policies in Reinforcement Learning @RLC-2024},
year={2024},
url={https://openreview.net/forum?id=FX7YtfEYj8},
note={\underline{Accepted as oral presentation}},
}


@inproceedings{sunny,
      title={Sometimes {I} am a Tree: Data drives fragile hierarchical generalization}, 
      author={Tian Qin and  Saphra, Naomi and David Alvarez-Melis},
      year={2024},
    booktitle={NeurIPS Workshop on Scientific Methods for Understanding Deep Learning \emph{and} NeurIPS Workshop on Compositional Learning},
    url = {https://arxiv.org/abs/2412.04619},
}

@inproceedings{sara-loss,
title={Loss in the Crowd:  Hidden Breakthroughs in Language Model Training},
author={Sara Kangaslahti and Elan Rosenfeld and Saphra, Naomi},
year={2024},
    booktitle={ICML Workshop on Mechanistic Interpretability},
note={Spotlighted},
url = {https://openreview.net/pdf/b0953a49ea69c118f2433f8bd34fcd5cf7c33bd4.pdf},
}

@inproceedings{attrib,
	keywords = {workshop},
	title = {Shapley Interactions for Complex Feature Attribution
},
	author = {Divyansh Singhvi* and Andrej Erkelens* and Raghav Jain* and Diganta Misra and  Saphra, Naomi},
    booktitle={NeurIPS Workshop on Attribution at Scale},
	year = {2023},
url = {https://arxiv.org/abs/2403.13106},
}

@inproceedings{rosie,
      title={Distributional Scaling Laws for Emergent Capabilities}, 
      author={Rosie Zhao and Sham Kakade and  Saphra, Naomi},
      year={2024},
    booktitle={NeurIPS Workshop on Scientific Methods for Understanding Deep Learning},
url = {https://openreview.net/forum?id=e8eo9iEFaO#discussion},
}

@inproceedings{jenny,
      title={Causation Does Not Imply Correlation: A Study of Circuit Mechanisms and Model Behaviors}, 
      author={Jenny Kaufmann* and Victoria R. Li* and Martin Wattenberg and David Alvarez-Melis and  Saphra, Naomi},
    booktitle={NeurIPS Workshop on Scientific Methods for Understanding Deep Learning},
      year={2024},
url = {https://openreview.net/forum?id=JYyqhr8zJ8#discussion},
}

@misc{prashanth2024recitereconstructrecollectmemorization,
      title={Recite, Reconstruct, Recollect: Memorization in {LMs} as a Multifaceted Phenomenon}, 
      author={USVSN Sai Prashanth* and Alvin Deng* and Kyle O'Brien* and Jyothir S V* and Mohammad Aflah Khan and Jaydeep Borkar and Christopher A. Choquette-Choo and Jacob Ray Fuehne and Stella Biderman and Tracy Ke and Katherine Lee and  Saphra, Naomi},
      year={2025},
      booktitle={{International Conference on Learning Representations (ICLR)}},
url = {https://arxiv.org/abs/2406.17746}
}

@misc{polypythias,
      title={PolyPythias: Stability and Outliers across Fifty Language Model Pre-Training Runs}, 
      author={Oskar van der Wal and Pietro Lesci and Max Müller-Eberstein and Naomi Saphra and Hailey Schoelkopf and Willem Zuidema and Stella Biderman},
      year={2025},
      booktitle={{International Conference on Learning Representations (ICLR)}},
url = {https://openreview.net/pdf/b10ca07b6fe02cd1810904d7fb5a4804254c0a64.pdf}
}

@misc{dynet,
  title={{DyNet}: The Dynamic Neural Network Toolkit},
  author={Graham Neubig and Chris Dyer and Yoav Goldberg and Austin Matthews and Waleed Ammar and Antonios Anastasopoulos and Miguel Ballesteros and David Chiang and Daniel Clothiaux and Trevor Cohn and Kevin Duh and Manaal Faruqui and Cynthia Gan and Dan Garrette and Yangfeng Ji and Lingpeng Kong and Adhiguna Kuncoro and Gaurav Kumar and Chaitanya Malaviya and Paul Michel and Yusuke Oda and Matthew Richardson and Naomi Saphra and Swabha Swayamdipta and Pengcheng Yin},
  journal={arXiv preprint},
  year={2017},
url = {https://arxiv.org/abs/1701.03980},
}



@inproceedings{sarah-mech,
title={Mechanistic?},
author={Saphra, Naomi and Sarah Wiegreffe*},
booktitle={EMNLP BlackboxNLP Workshop},
year={2024},
url={https://arxiv.org/abs/2410.09087},
note={Accepted as oral presentation},
}



@inproceedings{sultan,
	keywords = {workshop},
	title = {Towards out-of-distribution generalization in large-scale astronomical surveys: robust networks learn similar representations
},
	booktitle = {NeurIPS workshop on Machine Learning and the Physical Sciences},
	author = {Yash Gondhalekar and Sultan Hassan and  Saphra, Naomi and Sambatra Andrianomena},
	year = {2023},
url={https://arxiv.org/abs/2311.18007},
}


@inproceedings{zhao_one_2022,
	title = {One {Venue}, {Two} {Conferences}: {The} {Separation} of {Chinese} and {American} {Citation} {Networks}},
	shorttitle = {One {Venue}, {Two} {Conferences}},
	url = {https://openreview.net/forum?id=9U6w1tSZu_T},
	abstract = {At NeurIPS, American and Chinese institutions cite papers from each other's regions substantially less than they cite endogamously. We build a citation graph to quantify this divide, compare it to European connectivity, and discuss the causes and consequences of the separation.},
	language = {en},
	urldate = {2022-10-26},
	author = {Zhao*, Bingchen and Gu*, Yuling and Forde, Jessica Zosa and Saphra, Naomi},
	year = {2022},
	booktitle = {NeurIPS Workshop on Cultures of AI and AI for Culture},
}

@inproceedings{saphra_sparsity,
	keywords = {workshop},
	title = {Sparsity Emerges Naturally in Neural Language Models},
	booktitle = {ICML Workshop on Identifying and Understanding Deep Learning Phenomena (Deep Phenomena)},
	author = {Saphra, Naomi and Lopez, Adam},
	year = {2019},
url={https://arxiv.org/abs/1908.01817},
}

@inproceedings{kate,
	keywords = {workshop},
	title = {Carbon {AI} and the Concentration of Computational Work},
	booktitle = {Challenging the Work Society: an interdisciplinary summit},
	author = {McCurdy, Kate and Saphra, Naomi},
	year = {2019},
}

@inproceedings{pos_first,
	keywords = {workshop},
	title = {Evaluating {Informal}-{Domain} {Word} {Representations} {with} {UrbanDictionary}},
	booktitle = {ACL Workshop on {Evaluating} {Vector} {Space} {Representations} for {NLP} (RepEval)},
	author = {Saphra, Naomi and Lopez, Adam},
	year = {2016},
url={https://arxiv.org/abs/1606.08270},
}



@inproceedings{cotterell_algerian_2014,
	keywords = {workshop},
	title = {An {Algerian} {Arabic}-{French} {Code}-{Switched} {Corpus}},
	author = {Cotterell, Ryan and Renduchintala, Adithya and Saphra, Naomi and Callison-Burch, Chris},
    booktitle = {LREC Workshop on Free/Open-Source Arabic Corpora},
	year = {2014},
        note={Received best paper award},
url={https://www.cis.upenn.edu/~ccb/publications/arabic-french-codeswitching.pdf},
}

@inproceedings{schneider_framework_2014,
	keywords = {workshop},
	title = {A framework for (under) specifying dependency syntax without overloading annotators},
	booktitle = {ACL Linguistic {Annotation} {Workshop}},
	author = {Schneider, Nathan and O'Connor, Brendan and Saphra, Naomi and Bamman, David and Faruqui, Manaal and Smith, Noah A. and Dyer, Chris and Baldridge, Jason},
	year = {2014},
url={https://aclanthology.org/W13-2307.pdf},
}


@article{hu,
      title={Delays, Detours, and Forks in the Road: Latent State Models of Training Dynamics}, 
      author={Michael Hu and Angelica Chen and Naomi Saphra and Kyunghyun Cho},
      year={2023},
      journal={Transactions of Machine Learning Research (TMLR)},
url = {https://arxiv.org/abs/2308.09543},
}

@article{hupkes_state---art_2023,
	title = {State-of-the-art generalisation research in {NLP}: a taxonomy and review},
	author = {Hupkes, Dieuwke and Giulianelli, Mario and Dankers, Verna and Artetxe, Mikel and Elazar, Yanai and Pimentel, Tiago and Christodoulopoulos, Christos and Lasri, Karim and Saphra, Naomi and Sinclair, Arabella and Ulmer, Dennis and Schottmann, Florian and Batsuren, Khuyagbaatar and Sun, Kaiser and Sinha, Koustuv and Khalatbari, Leila and Ryskina, Maria and Frieske, Rita and Cotterell, Ryan and Jin, Zhijing},
	year = {2023},
	journal = {Nature Machine Intelligence},
url = {https://arxiv.org/abs/2210.03050},
}


@article{saphra2023interp,
    author = {Naomi Saphra},
    title = {Interpretability Creationism},
    journal = {The Gradient},
    year = {2023},
    url = {https://thegradient.pub/interpretability-creationism},
}

@inproceedings{zhang2024transcendence,
      title={Transcendence: Generative Models Can Outperform The Experts That Train Them}, 
      author={Edwin Zhang and Vincent Zhu and  Saphra, Naomi and Anat Kleiman and Benjamin L. Edelman and Milind Tambe and Sham M. Kakade and Eran Malach},
      year={2024},
      booktitle={Neural Information Processing Systems (NeurIPS)},
url={https://arxiv.org/abs/2406.11741},
}

@inproceedings{ff,
      title={Fast Forwarding Low-Rank Training}, 
      author={Adir Rahamim and Saphra, Naomi and Sara Kangaslahti and Yonatan Belinkov},
      year={2024},
      booktitle = {{Empirical Methods in Natural Language Processing (EMNLP)}},
url={https://arxiv.org/abs/2409.04206},
}


@inproceedings{li2024chatgptdoesnttrustchargers,
      title={{ChatGPT} Doesn't Trust {Chargers} Fans: Guardrail Sensitivity in Context}, 
      author={Victoria R. Li* and Yida Chen* and Saphra, Naomi},
      year={2024},
      booktitle = {{Empirical Methods in Natural Language Processing (EMNLP)}},
url={https://arxiv.org/abs/2407.06866},
}


@inproceedings{systematicity,
      title={Attribute Diversity Determines the Systematicity Gap in {VQA}}, 
      author={Ian Berlot-Attwell and Kumar Krishna Agrawal and A. Michael Carrell  and Yash Sharma and Naomi Saphra},
      year={2024},
      booktitle = {{Empirical Methods in Natural Language Processing (EMNLP)}},
url={https://aclanthology.org/2024.emnlp-main.537.pdf},
}


@inproceedings{bench,
      title={Benchmarks as Microscopes: A Call for Model Metrology}, 
      author={Michael Saxon and Ari Holtzman  and Peter West and William Yang Wang and Naomi Saphra},
      year={2024},
	booktitle = {{Conference on Language Modeling (COLM)}},
url={https://arxiv.org/abs/2407.16711},
}

@inproceedings{parse,
      title={First Tragedy, then Parse: History Repeats Itself in the New Era of Large Language Models}, 
      author={Naomi Saphra and Eve Fleisig and Kyunghyun Cho and Adam Lopez},
      year={2024},
	booktitle = {{North American Association for Computational Linguistics (NAACL)}},
url={https://arxiv.org/abs/2311.05020},
}



@inproceedings{ankner,
      title={Dynamic Masking Rate Schedules for {MLM} Pretraining}, 
      author={Zachary Ankner and Naomi Saphra and Davis Blalock and Jonathan Frankle and Matthew L. Leavitt},
      year={2024},
	booktitle = {{European Association for Computational Linguistics (EACL)}},
    note={Accepted as oral presentation},
url={https://arxiv.org/abs/2305.15096},
}


@inproceedings{chen,
      title={Sudden Drops in the Loss: Syntax Acquisition, Phase Transitions, and Simplicity Bias in {MLM}s}, 
      author={Angelica Chen and Ravid Schwartz-Ziv and Kyunghyun Cho and Matthew Leavitt and Naomi Saphra},
      year={2024},
	booktitle = {{International Conference on Learning Representations (ICLR)}},
    note={Spotlighted},
url={https://arxiv.org/abs/2309.07311},
}

@inproceedings{sherborne_tram_2023,
	title = {{TRAM}: {Bridging} {Trust} {Regions} and {Sharpness} {Aware} {Minimization}},
	year = {2024},
	author = {Sherborne, Tom and Naomi Saphra and Dasigi, Pradeep and Peng, Hao},
	booktitle = {{International Conference on Learning Representations (ICLR)}},
    note={Spotlighted},
    url = {https://arxiv.org/abs/2310.03646},
}

@inproceedings{juneja_linear_2023,
	title = {Linear {Connectivity} {Reveals} {Generalization} {Strategies}},
	author = {Juneja, Jeevesh and Bansal, Rachit and Cho, Kyunghyun and Sedoc, João and Saphra, Naomi},
	year = {2023},
	booktitle = {{International Conference on Learning Representations (ICLR)}},
url={https://arxiv.org/abs/2205.12411},
}


@inproceedings{sellam_multiberts_2021,
	keywords = {conference},
	title = {The {MultiBERTs}: {BERT} {Reproductions} for {Robustness} {Analysis}},
	shorttitle = {The {MultiBERTs}},
	author = {Sellam, Thibault and Yadlowsky, Steve and Tenney, Ian and Wei, Jason and Saphra, Naomi and D'Amour, Alexander and Linzen, Tal and Bastings, Jasmijn and Turc, Iulia Raluca and Eisenstein, Jacob and Das, Dipanjan and Pavlick, Ellie},
	year = {2022},
	booktitle = {{International Conference on Learning Representations (ICLR)}},
    note={Spotlighted},
url={https://arxiv.org/abs/2106.16163},
}

@inproceedings{white_nonlinear_2020,
	keywords = {conference},
	title = {A {Non-Linear} {Structural} {Probe}},
	shorttitle = {Pareto {Probing}},
	booktitle = {{North American Association for Computational Linguistics (NAACL)}},
	author = {White, Jennifer C. and Pimentel, Tiago and Naomi Saphra and Williams, Adina and Cotterell, Ryan},
	year = {2021},
url={https://arxiv.org/abs/2105.10185},
}

@inproceedings{saphra_lstms_2020,
	keywords = {conference},
	title = {{LSTMs} {Compose}---and {Learn}---{Bottom}-{Up}},
	booktitle = {Findings of {Empirical Methods in Natural Language Processing (EMNLP Findings)}},
	author = {Saphra, Naomi and Lopez, Adam},
	year = {2020},
url={https://arxiv.org/abs/2010.04650},
}

@inproceedings{pimentel_pareto_2020,
	keywords = {conference},
	title = {Pareto {Probing}: {Trading} {Off} {Accuracy} for {Complexity}},
	shorttitle = {Pareto {Probing}},
	booktitle = {{Empirical Methods in Natural Language Processing (EMNLP)}},
	author = {Pimentel*, Tiago and  Saphra*, Naomi and Williams, Adina and Cotterell, Ryan},
	year = {2020},
url={https://arxiv.org/abs/2010.02180},
}

@inproceedings{tahaei,
	keywords = {conference},
    author = {Tahaei, Mohammad and Vaniea, Kami and Saphra, Naomi},
    title = {Understanding Privacy-Related Questions on Stack Overflow},
    year = {2020},
    booktitle = {Conference on Human Factors in Computing Systems (CHI)},
url={https://dl.acm.org/doi/10.1145/3313831.3376768},
}



@inproceedings{saphra_understand_2019,
	keywords = {conference},
	title = {Understanding {Learning} {Dynamics} {Of} {Language} {Models} with {SVCCA}},
	booktitle = {{North American Association for Computational Linguistics (NAACL)}},
	author = {Saphra, Naomi and Lopez, Adam},
	year = {2019},
url={https://aclanthology.org/N19-1329/},
}


@inproceedings{naomi_saphra_amrica:_2015,
	keywords = {conference},
	title = {{AMRICA}: an {AMR} {Inspector} for {Cross}-language {Alignments}},
	booktitle = {{North American Association for Computational Linguistics (NAACL)} {(demos)}},
	author = {Saphra, Naomi and Lopez, Adam},
	year = {2015},
url={https://aclanthology.org/N15-3008/},
}


@inproceedings{vedaldi_understanding_2014,
	keywords = {conference},
	title = {Understanding {Objects} in {Detail} with {Fine}-grained {Attributes}},
	booktitle = {{Computer} {Vision} and {Pattern} {Recognition} ({CVPR})},
	url = {https://ieeexplore.ieee.org/document/6909858},
	author = {Vedaldi, Andrea and Mahendran, Siddharth and Tsogkas, Stavros and Maji, Subhransu and Girshick, Ross and Kannala, Juho and Rahtu, Esa and Kokkinos, Iasonas and Blaschko, Matthew B. and Weiss, David and Taskar, Ben and Simonyan, Karen and Saphra, Naomi and Mohamed, Sammy},
	year = {2014},
url={https://ieeexplore.ieee.org/document/6909858},
}

@inproceedings{valvoda_learning_2022,
	keywords = {conference},
	title = {Learning {Transductions} to {Test} {Systematic} {Compositionality}},
	booktitle = {{International Conference on Computational Linguistics (COLING)}},
	author = {Valvoda, Josef and Saphra, Naomi and Rawski, Jonathan and Cotterell, Ryan and Williams, Adina},
	year = {2022},
url={https://rycolab.io/papers/valvoda+al.coling22.pdf},
}