I'm an NLP researcher based in Pittsburgh. Currently, I'm working on retrieval augmented language models at Cohere. Before that I spent four years at Google, also working on retrieval augmented LMs. Most recently I was a Senior Research Scientist at Google DeepMind.

Before that, I spent a long time at UMass Amherst. I earned a PhD in computer science from UMass Amherst where I worked on knowledge representation and reasoning in the Information Extraction and Synthesis Laboratory (IESL) with Professor Andrew McCallum. My thesis work was on representing and automatically constructing embedded knowledge bases. While at UMass, I also earned a BS in Neuroscience, a BA and MS in Computer Science and a minor in Philosophy.

Selected Publications

• Attributed Question Answering: Evaluation and Modeling for Attributed Large Language Models
  Bernd Bohnet*, Vinh Q Tran*, Pat Verga*, et al.
  under review
• To Adapt or to Annotate: Challenges and Interventions for Domain Adaptation in Open-Domain Question Answering
  Dheeru Dua, Emma Strubell, Sameer Singh, Pat Verga
  ACL 2023
• Adaptable and Interpretable Neural Memory Over Symbolic Knowledge
  Pat Verga, Haitian Sun, Livio Baldini Soares, William Cohen
  NAACL 2021
• Simultaneously Self-attending to All Mentions for Full-Abstract Biological Relation Extraction
  Pat Verga , Emma Strubell, and Andrew McCallum.
  NAACL 2018. [code] [bibtex]
• [Best Long Paper Award] Linguistically-Informed Self-Attention for Semantic Role Labeling
  Emma Strubell, Pat Verga , Daniel Andor, David Weiss and Andrew McCallum.
  EMNLP 2018.
  [bibtex]

Other Things About Pat

• I'm originally from Gloucester, Massachusetts .
• I play bass and mandolin.
• I spend a lot of time both working on my Old House and watching This Old House.
• I like traveling, hiking, and camping and taking pictures of the places I go.
• I have two dogs, Nala and Pepper and previously had a free range bearded dragon named Bran (R.I.P.).
• Emma Strubell and I made a smart plant named Plant Jones - though he would have been slightly smarter if LLMs had existed at the time.