I'm a fifth year graduate student at the University of Washington in Computer Science and Engineering and am part time at the Allen Institute for Artificial Intelligence. I work with Yejin Choi and Ali Farhadi. My pronouns are he/his.
My research spans natural language processing, computer vision, and artificial intelligence. I'm excited about commonsense language understanding. As humans, our language is rooted in the complex world around us -- we use it to learn new concepts, share information, and collaborate with others. I'm interested in bridging the gap between what existing machine-learning approaches can do, and this humanlike level of language understanding grounded in the world. I'm also interested in exploring the social impacts of these technologies.
You can follow me on Twitter at @rown.
We introduce MERLOT, a model that learns multimodal script knowledge by watching millions of YouTube videos with transcribed speech -- in an entirely label-free, self-supervised manner. Our model not only learns to match images to temporally corresponding words, but also to contextualize what is happening globally over time. When finetuned it gets SOTA on 12 datasets requiring visual (and temporal) reasoning.
PIGLeT: Language Grounding Through Neuro-Symbolic Interaction in a 3D World
Rowan Zellers, Ari Holtzman, Matthew Peters, Roozbeh Mottaghi, Aniruddha Kembhavi, Ali Farhadi, Yejin Choi. ACL 2021. [paper] [project page]
We introduce a model named PIGLeT that learns physical commonsense understanding by interacting with the world, and uses this knowledge to ground language. This strategy works much better than modeling everything solely through linguistic form. When forecasting "what happens next" given an English sentence, our base-sized PIGLeT model outperforms text-to-text models that are 100x larger.
Big transformers do well on NLP benchmarks, but there's a gap between today's benchmarks and how humans use language. We narrow this gap, instead evaluating machines dynamically by their real-world language use. Rather than bubbling the right answers on multiple choice tests, the idea is to write text that's helpful to people in need. Today's models show room for improvement in this setting.
Constraints like "generate a recipe using the following ingredients" are tricky for language models to follow, even after finetuning. We introduce NeuroLogic Decoding, an algorithm that allows for the generation of high-quality text while satisfying constraints, allowing us to avoid the finetuning step completely.
Today's best language understanding models are trained on massive amounts of purely textual data. We show that their representations have structural similarity to models trained on visual data, though this cross-modal linkage is far from perfect.
Many images online are edited in some way, but it is the intent that separates harmful edits like deepfakes from innocuous edits like an enhanced vacation photo. We introduce a new task and dataset for reasoning about why an image was edited, and a new model for the task.
Today's models achieve superhuman performance on benchmarks like ImageNet and Stanford Natural Language Inference (SNLI), but it is unclear whether they solved the underlying task, or rather overfitted to dataset biases. We study AFLite, an extension of Adversarial Filtering, to remove dataset biases. Filtering datasets like ImageNet makes them much harder for machines, while human performance remains high.
To apply eyeshadow without a brush, should I use a cotton swab or a toothpick? We introduce a benchmark for evaluating the ability for machines to reason about physical situations through natural language. Today's pretrained language models struggle, showing room for future reesearch.
Can adversaries use state-of-the-art NLP models to generate "neural fake news"? We investigate the threat of machine-written propaganda that mimics the style of real news, through a model named Grover. Given a headline like `Link Found Between Vaccines and Autism,' Grover can generate the rest of the article. The best defense agianst Grover turns out to be Grover itself, demonstrating the importance of public release of strong generators.
We show that commonsense inference still proves difficult for even state-of-the-art models, by presenting HellaSwag, a new challenge dataset. Though its questions are trivial for humans, even deep pretrained models (like BERT) struggle. The key insight is to scale up the length and complexity of the dataset examples towards a critical zone wherein generated text is ridiculous to humans, yet often misclassified by state-of-the-art models.
We formulate the new task of Visual Commonsense reasoning, where a model must not only answer challenging visual questions expressed in natural language: it must provide a rationale explaining why its answer is true. We introduce a new dataset, VCR, consisting of 290k multiple choice QA problems derived from 110k movie scenes.
Press: New York Times
We release a new NLI dataset, SWAG, with 113k challenging multiple choice questions about grounded sitautions. To build this dataset, we present Adversarial Filtering (AF), which allows for data collection at scale while minimizing annotation artifacts.
We study scene graph generation: building a graph where the nodes are objects and the edges are pairwise relationships. The visual world has many repeating structures (motifs). We built a model to capture them, which improves significantly over the prior staet-of-the-art.
We investigate zero-shot multimodal learning where the topics of classification are verbs, not objects. We crowdsource verb attributes and build a model to learn them from unlabeled text, and dictionary definitions. We used these verb attributes to recognize actions in images.
Multimodal Sentiment Intensity Analysis in Videos: Facial Gestures and Verbal Messages
Amir Zadeh, Rowan Zellers, Eli Pincus, Louis-Philippe Morency. IEEE Intelligent Systems 2016 [paper]
We provide a study of sentiment analysis applied on video data, not just text. We present a model that exploits the dynamics between gestures and verbal messages.
Nucleotide Interdependency in Transcription Factor Binding Sites in the Drosophila Genome
Jacqueline Dresch, Rowan Zellers, Daniel Bork, Robert Drewell. Gene Regulation and Systems Bio 2016. [paper]
MARZ: an algorithm to combinatorially analyze gapped n-mer models of transcription factor
Rowan Zellers, Robert Drewell, Jacqueline Dresch. BMC Bioinformatics 2015. [paper+code]
We model the specificity with which regulatory proteins bind to DNA sequences during embryonic development. We use this model to study binding sites for 15 distinct regulatory proteins in the Drosophila (fruit fly) genome.
I help advise several talented undergrads on research:
My email address is rowanz at cs.washington.edu. Note: