Yufei Li Email & Phone Number

Menlo Park, Ca, Us

Riverside, California, United States

Researching and developing ML-assisted interactive systems at the intersection of NLP, model uncertainty and robustness, effective and efficient training/inference with Professor Cong Liu.Publications:ACL Findings 2024 (First): https://arxiv.org/pdf/2310.04994ACL 2024 (Second): https://arxiv.org/pdf/2312.06924RTSS 2023 (First): https://arxiv.org/pdf/2309.06619.pdfRTSS 2023 (Third): https://arxiv.org/pdf/2308.15039.pdfACL 2023 (First):… Show more Researching and developing ML-assisted interactive systems at the intersection of NLP, model uncertainty and robustness, effective and efficient training/inference with Professor Cong Liu.Publications:ACL Findings 2024 (First): https://arxiv.org/pdf/2310.04994ACL 2024 (Second): https://arxiv.org/pdf/2312.06924RTSS 2023 (First): https://arxiv.org/pdf/2309.06619.pdfRTSS 2023 (Third): https://arxiv.org/pdf/2308.15039.pdfACL 2023 (First): https://aclanthology.org/2023.acl-short.116.pdfACL 2023 (First): https://aclanthology.org/2023.acl-long.100.pdfEMNLP 2022 (Second): https://aclanthology.org/2022.emnlp-main.761.pdf Show less

Mountain View, California, United States

- Generated a large-scale synthetic rewrite benchmark (including factuality, stylish, and conversational) using LLM with hand-crafted prompts. - Instruction-tuned (SFT) a model on the rewrite benchmark, and distilled reward models from LLM preferences on generated model responses. - RLHF using the decoupled reward models (PPO), targeting instruction agreement, rewrite coherence, and surgical difference with the original response.- Evaluated rewrite quality through prompting… Show more - Generated a large-scale synthetic rewrite benchmark (including factuality, stylish, and conversational) using LLM with hand-crafted prompts. - Instruction-tuned (SFT) a model on the rewrite benchmark, and distilled reward models from LLM preferences on generated model responses. - RLHF using the decoupled reward models (PPO), targeting instruction agreement, rewrite coherence, and surgical difference with the original response.- Evaluated rewrite quality through prompting AutoRaters (e.g., Gemini, Bard). Show less

Princeton, New Jersey, United States

- Annotated fields (named entities) in log messages, and leveraged BART for prompt-based field extraction.- Defined hierarchical relations between log components and configured dynamic attributed graphs.- Detected anomalies in log messages using a GNN-based encoder enhanced with temporal-attentive transformers.Publications:ICKG 2023 (First): https://arxiv.org/pdf/2309.05953.pdf

Dallas, Texas, United States

Work @ Dr. Wei Yang's Lab, researching natural language processing and software engineering.Publications:ESEC/FSE 2021 (Co-First): https://arxiv.org/pdf/2106.10507.pdf

Princeton, New Jersey, United States

- Annotated name entities and relations using regular expression rules in CVE corpus for distant supervision.- Incorporated pre-trained GPT-2 into a sequence labeling framework for information extraction (IE).- Proposed a bootstrap training strategy for denoising distant labels and selecting high-quality instances.

Beijing, China

- Built a joint key point and pose recognition model for character identification tasks.- Developed an adaptive discrimination definition mode from Caffe to TensorFlow.- Designed a lightweight CNN to calibrate video frame orientations in real-time for online streaming.

Doctor Of Philosophy - Phd, Computer Engineering, 4.0/4.0

I'm a current PhD student advised by Dr. Cong Liu, doing research at natural language processing applications and efficient language model.

Master Of Science - Ms, Computer Engineering

Glad to do research at Julian McAuley's Lab, collaborating with Shuyang Li and Zhenkui He.

Bachelor Of Science - Bs, Mechanical Engineering

Activities and Societies: Badminton, Student Union, VEX Robotics Competition Team Leader & Programmer (representing XJTU) in VEX Robotics.