Scholar
Jiao Sun
Google Scholar ID: 7xWN15gAAAAJ
Google DeepMind
Citations & Impact
Citations
5,419
H-index
19
i10-index
25
Publications
20
Co-authors
7
list available
Publications
Resume (English only)
Academic Achievements
- ACL 2023: "Dialect-robust Evaluation of Generated Text"
- NeurIPS 2023: Co-authored "LIMA: Less is More for Alignment"
- EMNLP 2022: Published multiple papers including "Controlled Pun Generation", "Pun with Explanation", "Investigate the Benefits of Free-form Rationales", and "Robustness of Bias Evaluation"
- EMNLP 2021: "Paraphrase Generation" (papers AESOP and ESTER)
- ACL 2021: Work on event bias in Wikipedia received Best Paper Nomination
- CHI 2022: Research on bias in greeting cards received Best Paper Honorable Mention
- Selected as a 2023 EECS Rising Star
Background
- Ph.D. candidate in Natural Language Processing at the University of Southern California (USC)
- Amazon Fellow
- Research focuses on trustworthy text generation
- Specific interests include controlled text generation, robustness of NLG systems, NLG evaluation, and data efficiency
- Actively involved in large language model (LLM) research, including pretraining on 60+ TB data, fine-tuning mT5 models up to XXL scale, distilling PaLM's dialect rewriting capability into EdiT5, and co-developing LIMA—a high-quality LLM trained on only 1,000 examples
- Advocates for fair AI; studied bias in Wikipedia event descriptions and greeting cards
