John
Wieting

John Wieting

Senior Research Scientist
Google DeepMind, New York
Ph.D., Carnegie Mellon University, 2020

I am a senior research scientist at Google DeepMind in New York. I work on natural language processing — for most of the last decade on learning representations of meaning, and more recently on watermarking language models and on what makes long-form generated text good or bad. I have also contributed to the Gemini, Gemma, and PaLM model families.

I did my Ph.D. at Carnegie Mellon, advised by Taylor Berg-Kirkpatrick and Graham Neubig, working closely with Kevin Gimpel at TTIC. Before that I did an MS with Dan Roth at Illinois, and undergraduate degrees in mathematics and chemistry at Wisconsin.

The best way to reach me is by email.

News

  • New preprint: POLARIS, on guiding small models to write long stories.
  • New preprint: StoryScope, investigating idiosyncrasies in AI fiction.
  • Two watermarking papers accepted to TMLR.

Publications

  1. 51 · 2026

    POLARIS: Guiding Small Models to Write Long Stories

    Rishanth Rajendhran, Jenna Russell, Mohit Iyyer, John Wieting Preprint

  2. 50 · 2026

    StoryScope: Investigating Idiosyncrasies in AI Fiction

    Jenna Russell, Rishanth Rajendhran, Chau Minh Pham, Mohit Iyyer, John Wieting Preprint

  3. 49 · 2026

    Improving Detection of Watermarked Language Models

    Dara Bahri, John Wieting TMLR 2026

    Tackles the detection side of watermarking, showing that combining watermark-based detection with model-based detectors outperforms either alone — including on text from unwatermarked models.

  4. 48 · 2026

    A Watermark for Black-Box Language Models

    Dara Bahri, John Wieting TMLR 2026

    A watermarking scheme requiring only sampling access to a model — no logits — making watermarking practical for API-only settings while retaining distortion-free guarantees.

  5. 47 · 2026

    Gemma 4 Technical Report

    Gemma Team 322 Google DeepMind · technical report

  6. 46 · 2025

    A Gold Standard Dataset for the Reviewer Assignment Problem

    Ivan Stelmakh, John Wieting, Yang Xi, Graham Neubig, Nihar B. Shah TMLR 2025

  7. 45 · 2025

    Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilities

    Gemini Team 3298 Google DeepMind · technical report

  8. 44 · 2025

    Gemma 3 Technical Report

    Gemma Team Google DeepMind · technical report

  9. 43 · 2024

    PostMark: A Robust Blackbox Watermark for Large Language Models

    Yapei Chang, Kalpesh Krishna, Amir Houmansadr, John Wieting, Mohit Iyyer EMNLP 2024

  10. 42 · 2024

    Leveraging LLMs for Synthesizing Training Data Across Many Languages in Multilingual Dense Retrieval

    Nandan Thakur, Jianmo Ni, Gustavo Hernandez Abrego, John Wieting, Jimmy Lin, Daniel Cer NAACL 2024

  11. 41 · 2024

    Multiple References with Meaningful Variations Improve Literary Machine Translation

    Si Wu, John Wieting, David A. Smith Preprint

  12. 40 · 2024

    Gemini 1.5: Unlocking multimodal understanding across millions of tokens of context

    Gemini Team 1136 Google DeepMind · technical report

  13. 39 · 2023

    Beyond Contrastive Learning: A Variational Generative Model for Multilingual Retrieval

    John Wieting, Jonathan Clark, William Cohen, Graham Neubig, Taylor Berg-Kirkpatrick ACL 2023

  14. 38 · 2023

    Paraphrasing evades detectors of AI-generated text, but retrieval is an effective defense

    Kalpesh Krishna, Yixiao Song, Marzena Karpinska, John Wieting, Mohit Iyyer NeurIPS 2023

    Showed that a discourse-level paraphraser (DIPPER) evades watermarking, GPTZero, DetectGPT, and OpenAI's own classifier — and that retrieval over previously generated text is the one defense that holds up. Widely cited in the debate over whether AI-text detection is possible at all.

  15. 37 · 2023

    Evaluating and Modeling Attribution for Cross-Lingual Question Answering

    Benjamin Muller, John Wieting, Jonathan H. Clark … 9 EMNLP 2023Benjamin Muller, John Wieting, Jonathan H. Clark, Tom Kwiatkowski, Sebastian Ruder, Livio Baldini Soares, Roee Aharoni, Jonathan Herzig, Xinyi Wang
  16. 36 · 2023

    Augmenting Pre-trained Language Models with QA-Memory for Open-Domain Question Answering

    Wenhu Chen, Pat Verga, Michiel de Jong, John Wieting, William W. Cohen EACL 2023

  17. 35 · 2023

    QA Is the New KR: Question-Answer Pairs as Knowledge Bases

    William W. Cohen, Wenhu Chen … Pat Verga, John Wieting 7 AAAI 2023William W. Cohen, Wenhu Chen, Michiel de Jong, Nitish Gupta, Alessandro Presta, Pat Verga, John Wieting
  18. 34 · 2023

    FIAT: Fusing Learning Paradigms with Instruction-Accelerated Tuning

    Xinyi Wang, John Wieting, Jonathan H. Clark Preprint, 2023

    Fused the two dominant learning paradigms — in-context instruction and parameter tuning — into a single method that outperforms both ICL and fine-tuning alone across model scales.

  19. 33 · 2023

    XTREME-UP: A User-Centric Scarce-Data Benchmark for Under-Represented Languages

    Sebastian Ruder, Jonathan H. Clark … John Wieting … Partha Talukdar 27 Findings of EMNLP 2023Sebastian Ruder, Jonathan H. Clark, Alexander Gutkin, Mihir Kale, Min Ma, Massimo Nicosia, Shruti Rijhwani, Parker Riley, Jean-Michel A. Sarr, Xinyi Wang, John Wieting, Nitish Gupta, Anna Katanova, Christo Kirov, Dana L. Dickinson, Brian Roark, Bidisha Samanta, Connie Tao, David I. Adelani, Vera Axelrod, Isaac Caswell, Colin Cherry, Dan Garrette, Reeve Ingle, Melvin Johnson, Dmitry Panteleev, Partha Talukdar
  20. 32 · 2023

    Evaluating Large Language Models on Controlled Generation Tasks

    Jiao Sun, Yufei Tian … John Wieting … Xuezhe Ma 9 EMNLP 2023Jiao Sun, Yufei Tian, Wangchunshu Zhou, Nan Xu, Qian Hu, Rahul Gupta, John Wieting, Nanyun Peng, Xuezhe Ma

    A systematic audit of LLMs on controlled generation, finding they still fail at fine-grained constraints — such as syntax — that smaller, specialized models handle.

  21. 31 · 2023

    Gemini: A Family of Highly Capable Multimodal Models

    Gemini Team 1350 Google DeepMind · technical report

  22. 30 · 2023

    PaLM 2 Technical Report

    Rohan Anil, Andrew M. Dai … John Wieting … Yonghui Wu 128 Google · technical report

  23. 29 · 2022

    Canine: Pre-training an Efficient Tokenization-Free Encoder for Language Representation

    Jonathan H. Clark, Dan Garrette, Iulia Turc, John Wieting TACL 2022

    The first pre-trained tokenization-free encoder: CANINE operates directly on characters, sidestepping the brittleness of fixed subword vocabularies while outperforming a comparable mBERT model with fewer parameters.

  24. 28 · 2022

    RankGen: Improving Text Generation with Large Ranking Models

    Kalpesh Krishna, Yapei Chang, John Wieting, Mohit Iyyer EMNLP 2022

    A 1.2B-parameter ranker that scores how well candidate continuations follow a prefix; plugged into decoding, it substantially improves coherence over nucleus sampling in both automatic and human evaluations.

  25. 27 · 2022

    Paraphrastic Representations at Scale

    John Wieting, Kevin Gimpel, Graham Neubig, Taylor Berg-Kirkpatrick EMNLP 2022 (Demo)

    Scaled the simple-embeddings recipe to massive bitext in eight languages; the resulting models remain state of the art for unsupervised semantic similarity while running orders of magnitude faster on CPU than transformer encoders.

  26. 26 · 2022

    Exploring Document-Level Literary Machine Translation with Parallel Paragraphs from World Literature

    Katherine Thai, Marzena Karpinska … John Wieting, Mohit Iyyer 7 EMNLP 2022Katherine Thai, Marzena Karpinska, Kalpesh Krishna, Bill Ray, Moira Inghilleri, John Wieting, Mohit Iyyer
  27. 25 · 2022

    On the Ingredients of an Effective Zero-shot Semantic Parser

    Pengcheng Yin, John Wieting, Avirup Sil, Graham Neubig ACL 2022

  28. 24 · 2022

    Faithful to the Document or to the World? Mitigating Hallucinations via Entity-Linked Knowledge in Abstractive Summarization

    Yue Dong, John Wieting, Pat Verga Findings of EMNLP 2022

  29. 23 · 2021

    On Learning Text Style Transfer with Direct Rewards

    Yixin Liu, Graham Neubig, John Wieting NAACL 2021

  30. 22 · 2021

    Improving the Diversity of Unsupervised Paraphrasing with Embedding Outputs

    Monisha Jegadeesan, Sachin Kumar, John Wieting, Yulia Tsvetkov MRL Workshop 2021

  31. 21 · 2020

    A Bilingual Generative Transformer for Semantic Sentence Embedding

    John Wieting, Graham Neubig, Taylor Berg-Kirkpatrick EMNLP 2020

  32. 20 · 2020

    Reformulating Unsupervised Style Transfer as Paraphrase Generation

    Kalpesh Krishna, John Wieting, Mohit Iyyer EMNLP 2020

    Reframed unsupervised style transfer as paraphrase generation (STRAP): normalize the style away with a paraphraser, then style it back. No parallel data or style-specific engineering, with large gains over prior systems across formality, Shakespeare, and social-media styles.

  33. 19 · 2020

    Improving Candidate Generation for Low-resource Cross-lingual Entity Linking

    Shuyan Zhou, Shruti Rijhwani, John Wieting, Jaime Carbonell, Graham Neubig TACL 2020

  34. 18 · 2020

    Learning and Applications of Paraphrastic Representations for Natural Language

    John Wieting Ph.D. thesis, Carnegie Mellon University

  35. 17 · 2019

    Simple and Effective Paraphrastic Similarity from Parallel Translations

    John Wieting, Kevin Gimpel, Graham Neubig, Taylor Berg-Kirkpatrick ACL 2019

  36. 16 · 2019

    Beyond BLEU: Training Neural Machine Translation with Semantic Similarity

    John Wieting, Taylor Berg-Kirkpatrick, Kevin Gimpel, Graham Neubig ACL 2019

    Trained machine translation systems against a semantic-similarity reward rather than BLEU, improving translations by optimizing what evaluation was supposed to measure all along.

  37. 15 · 2019

    No Training Required: Exploring Random Encoders for Sentence Classification

    John Wieting, Douwe Kiela ICLR 2019

  38. 14 · 2019

    compare-mt: A Tool for Holistic Comparison of Language Generation Systems

    Graham Neubig, Zi-Yi Dou … Xinyi Wang, John Wieting 7 PreprintGraham Neubig, Zi-Yi Dou, Junjie Hu, Paul Michel, Danish Pruthi, Xinyi Wang, John Wieting

    Turns “system A beats system B by 0.7 BLEU” into an actual analysis: compare-mt breaks down where two generation systems differ — by bucket, n-gram, and individual sentence.

  39. 13 · 2018

    ParaNMT-50M: Pushing the Limits of Paraphrastic Sentence Embeddings with Millions of Machine Translations

    John Wieting, Kevin Gimpel ACL 2018

    Showed that back-translating a bitext yields tens of millions of paraphrase pairs essentially for free, and that simple embeddings trained on them outperform far more sophisticated architectures — a baseline the Sentence-BERT paper later acknowledged as remarkably strong. The dataset became a standard resource for paraphrase generation and semantic similarity.

  40. 12 · 2018

    Adversarial Example Generation with Syntactically Controlled Paraphrase Networks

    Mohit Iyyer, John Wieting, Kevin Gimpel, Luke Zettlemoyer NAACL 2018

    Introduced syntactically controlled paraphrase networks, which rewrite a sentence into a specified syntactic form — and showed these adversarial rephrasings break models that word-level perturbations leave intact.

  41. 11 · 2018

    Quality Signals in Generated Stories

    Manasvi Sagarkar, John Wieting, Lifu Tu, Kevin Gimpel *SEM 2018

  42. 10 · 2018

    CogCompNLP: Your Swiss Army Knife for NLP

    Daniel Khashabi, Mark Sammons … John Wieting … Dan Roth 23 LREC 2018Daniel Khashabi, Mark Sammons, Ben Zhou, Tom Redman, Christos Christodoulopoulos, Vivek Srikumar, Nicholas Rizzolo, Lev Ratinov, Guanheng Luo, Quang Do, Chen-Tse Tsai, Subhro Roy, Stephen Mayhew, Zhili Feng, John Wieting, Xiaodong Yu, Yangqiu Song, Shashank Gupta, Shyam Upadhyay, Naveen Arivazhagan, Qiang Ning, Shaoshi Ling, Dan Roth
  43. 9 · 2017

    Revisiting Recurrent Networks for Paraphrastic Sentence Embeddings

    John Wieting, Kevin Gimpel ACL 2017

  44. 8 · 2017

    Learning Paraphrastic Sentence Embeddings from Back-Translated Bitext

    John Wieting, Jonathan Mallinson, Kevin Gimpel EMNLP 2017

  45. 7 · 2016

    Towards Universal Paraphrastic Sentence Embeddings

    John Wieting, Mohit Bansal, Kevin Gimpel, Karen Livescu ICLR 2016 (oral)

    The first work to evaluate sentence embeddings directly on semantic textual similarity — a protocol later adopted by SentEval and its successors — it found that simply averaging word embeddings beat LSTMs for general-purpose sentence representation. This result became the starting point for Arora et al.'s (2017) famous “tough-to-beat baseline.”

  46. 6 · 2016

    Charagram: Embedding Words and Sentences via Character n-grams

    John Wieting, Mohit Bansal, Kevin Gimpel, Karen Livescu EMNLP 2016

    Posted to arXiv days before fastText's subword paper, Charagram independently made the same core argument: character n-grams, embedded simply, beat character LSTMs and CNNs for word and sentence representation.

  47. 5 · 2016

    UMD-TTIC-UW at SemEval-2016 Task 1: Attention-Based Multi-Perspective Convolutional Neural Networks for Textual Similarity

    Hua He, John Wieting, Kevin Gimpel, Jinfeng Rao, Jimmy Lin SemEval 2016

  48. 4 · 2015

    From Paraphrase Database to Compositional Paraphrase Model and Back

    John Wieting, Mohit Bansal, Kevin Gimpel, Karen Livescu TACL 2015

    The Paragram vectors trained here — including the 300d Paragram-SL999 set released in the appendix — were the first to reach human-level agreement on SimLex-999, and became a standard initialization for later similarity models.

  49. 3 · 2015

    Clustering With Side Information: From a Probabilistic Model to a Deterministic Algorithm

    Daniel Khashabi, Jeffrey Yufei Liu, John Wieting, Feng Liang Preprint

  50. 2 · 2014

    Tiered Clustering to Improve Lexical Entailment

    John Wieting Preprint

    A sole-authored early paper: tiered clustering over distributional contexts to sharpen lexical entailment detection.

  51. 1 · 2013

    Illinois Cognitive Computation Group UI-CCG TAC 2013 Entity Linking and Slot Filler Validation Systems

    Xiao Cheng, Bingling Chen … John Wieting … Dan Roth 10 TAC 2013Xiao Cheng, Bingling Chen, Rajhans Samdani, Kai-Wei Chang, Zhiye Fei, Mark Sammons, John Wieting, Subhro Roy, Chizheng Wang, Dan Roth

Software & data

  1. data

    ParaNMT-50M

    50 million English paraphrase pairs, produced by back-translating the Czech side of CzEng, with trained sentence embedding models. from ACL 2018

  2. models

    Paraphrastic representations at scale

    Trained paraphrastic sentence embedding models for English and 90+ languages, with a small interface for embedding text. from EMNLP 2022

  3. tool

    compare-mt

    Holistic comparison and analysis of the outputs of language generation systems. from arXiv:1903.07926

Service

Area Chair

  • AAAI2022, 2023, 2026 · Senior Program Committee, 2027
  • NeurIPS2023, 2024, 2025
  • ICML2025
  • ACL2020, 2024
  • NAACL2021, 2024
  • EMNLP2024
  • EACL2024
  • ARRongoing

Reviewing

  • TMLRongoing

Other

  • ACL 2020Paper Matching System — designed and trained the abstract similarity model used to assign reviewers to papers, and helped build the system with Graham Neubig

Invited talks

  • Georgia Tech — CS 4650, Natural Language Understanding
  • Google — NLX invited talk · Google — Lux Leads meeting
  • Ohio State University · UMass Amherst · Allen Institute for AI · IBM
  • Midwest Speech and Language Days, University of Chicago
  • Midwest Speech and Language Days, University of Chicago · CCICADA Research Retreat, Carnegie Mellon