Siyi Liu

About

I'm third year PhD student at Department of Computer and Information Science, University of Pennsylvania, advised by Prof. Dan Roth

My research interest lies within the areas of natural language processing. Recently, I’m interested in studying various types of conflicts in the applications of Large Language Models (LLMs), including

• hallucinations and alignment - LLM generations that are non-factual, toxic, or unsubstatiated/contradictory to the contexts [1]
• retrieval, knowledge conflicts, and reasoning - How can LLMs reason through conflicting knowledge from different sources (contextual/parametric knowledge)? [2]
• perspectives and biases - conflicts across different perspectives and opinions [5] [6] [7]

• Teaching LLMs to learn from failures/feedback
• Deepfake Video Detection
• Representation Learning for Authorship Representation

• Open Source Project (Core Contributor) - Distilled Generative Text Embedding Model
• Framing Bias Detection

Selected Publications

[1] Towards Long Context Hallucination Detection
Siyi Liu , Kishaloy Halder, et al.
in submission

[2] Open Domain Question Answering with Conflicting Contexts
Siyi Liu , Qiang Ning, et al.
in submission

[3] Using LLM for improving key event discovery: Temporal-guided news stream clustering with event summaries
Nishanth Nakshatri, Siyi Liu, Sihao Chen, Daniel Hopkins, Dan Roth, Dan Goldwasser
EMNLP 2023

[4] Open-Domain Event Graph Induction for Mitigating Framing Bias
Siyi Liu, Hongming Zhang, Hongwei Wang, Kaiqiang Song, Dan Roth, Dong Yu
arXiv

[5] Design Challenges for a Multi-Perspective Search Engine
Sihao Chen*, Siyi Liu* , Xander Uyttendaele, Yi Zhang, William Bruno, Dan Roth
NAACL 2022 Findings

[6] MultiOpEd: A Corpus of Multi-Perspective News Editorials
Siyi Liu, Sihao Chen, Xander Uyttendaele, Dan Roth
NAACL 2021
[Code] [Slides] [Poster] [Talk]

[7] Detecting frames in news headlines and its application to analyzing news framing trends surrounding US gun violence
Siyi Liu, Lei Guo, Kate Mays, Margrit Betke, Derry Tanti Wijaya
CoNLL 2019

[8] Learning to mirror speaking styles incrementally
Siyi Liu*, Ziang Leng*, Derry Wijaya
arXiv

Work Experience

AWS AI Labs

Applied Scientist Intern
Host: Kishaloy Halder May 2024 - Aug 2024

AWS AI Labs

Applied Scientist Intern
Host: Qiang Ning May 2023 - Aug 2023

Tencent AI Labs

NLP Summer Research Intern
Host: Hongming Zhang May 2022 - Aug 2022

Past Projects

Open Domain QA with Conflicting Contexts

25% of unambiguous, open domain questions can lead to conflicting contexts when retrieved using Google Search.

Information Pollution

Perspectives-oriented Search

Demo Website

An example screenshot of our Multi-Perspective Search Engine

A survey that compares the search results of our system and Google Search

MultiOpEd: A Corpus of Multi-Perspective News Editorials

News Framing

Detecting frames in news headlines and its application to analyzing news framing trends surrounding US gun violence