Experience

Mar. 2023 – May 2023

Conducted NLP (Natural Language Processing) research evaluating ChatGPT on 21 diverse datasets, covering various downstream tasks such as question answering, summarization, ranking, and sentiment analysis
Implemented prompt design to facilitate in-context learning (zero-shot and few-shot) and enhanced performance
Introduce an instruction engineering strategy that improves open LLMs (Large Language Models) to be competitive with closed-source large models across a set of challenging NLP tasks without any additional training; combine the strategy with LLaMA to reach competitiveness with OpenAI ChatGPT

Aug. 2022 – Apr. 2023

Conducted NLP (Natural Language Processing) research on scam detection in DeFi (decentralized finance)
Developed techniques including CodeBERT (a pre-trained model for both natural language and programming language) and other machine learning (deep learning) models to detect potential scams and to protect users from security vulnerabilities in smart contracts implemented by Solidity, with accuracy in this stage achieving 95.4%
Proposed zero-shot and few-shot classifiers with BLOOM (a large language model) and GPT-3; designed prompt tuning and Chain-of-Thought prompting strategies to improve performance, with F1 Score achieving 0.74