Erica Cai
I am a research scientist at FAIR, Meta Superintelligence Labs, working on RL post-training for image-to-code, OCR, and chart reasoning capabilities. I completed my PhD in Computer Science at the University of Massachusetts Amherst in May 2025 and was advised by Prof. Brendan T. O'Connor. I was a member of UMass NLP and the Statistical Social Language Analysis lab. My work on knowledge graphs, natural language processing, computational social science, causal structural learning, and causal inference is in Research.
From Fall 2023 through Spring 2025, I worked part-time at MIT Lincoln Laboratory, where I also interned during the summers. Prior to graduate school, I completed my undergraduate studies in Computer Science at Rutgers University in 2020, where I worked with Prof. Eric Allender in computational complexity theory and with Prof. Janne Lindqvist in developing methods to explore the privacy and security of ephemeral messages.
Recent News
- June, 2026: Traveling to Boston to give a talk on UniqueRank, which was accepted to NetSci and IEEE Transactions on Network Science and Engineering.
- March, 2026: AffilKG accepted at Language Resources and Evaluation Conference.
- December, 2025: Paper on error propagation of knowledge graph extraction on downstream graph analyses accepted at Applied Network Science.
- June, 2025: Joined Meta AI as a research scientist.
- May, 2025: Defended my PhD.
- March, 2025: FAME accepted at AAAI ICWSM.
- July, 2024: Paper accepted at Journal of Law and Courts; attended the Graph Exploitation Symposium in MIT Endicott House.
- May, 2024: Paper accepted at ACL (Findings); started my second summer research internship at MIT Lincoln Laboratory.
- April, 2024: Gave a talk at International Conference of Complex Networks in Exeter (UK) on low-resource information extraction and identification of influential nodes in knowledge graphs. The paper was published as part of Complex Networks XV, in Springer Nature.
- December, 2023: Presented our poster on a new zero-shot event extraction method at the NeurIPS Instruction Tuning and Instruction Following Workshop in New Orleans.
- November, 2023: Presented our poster on a new zero-shot event extraction method at Text as Data in Amherst.
- August, 2023: Presented our poster at the Graph Exploitation Symposium in MIT Endicott House on targeted few-shot fine-grained NER and relation extraction.
- July, 2023: Travelled to ACL in Toronto to present our paper on evaluating zero-shot event extraction methods and gave a talk about my event extraction research at MIT Lincoln Laboratory.
- May, 2023: Started my internship at MIT Lincoln Laboratory broadly on information extraction for constructing knowledge graphs.
- April, 2023: Gave a talk and presented our poster on zero-shot event extraction at New England NLP.
- March, 2023: Strict Scrutiny, a podcast about the Supreme Court, mentioned our paper.
In my free time, I play tennis and play the piano, and like to travel.
