Projects

Advancing LLM interpretability and AI ethics.

Current Projects

Explainable Censorship

Analyzing open-source moderation models to examine how they behave in ambiguous cases. Explores interpretability, calibration, and explanation techniques to improve transparency in automated content moderation using the UC Berkeley Measuring Hate Speech dataset.

Research Leads: Yiping Chen

Explainable Censorship
Explainable Censorship
Explainable Censorship
Explainable Censorship
Explainable Censorship
Content ModerationInterpretabilityFairnessTransformersNLP

Context Based Captioning

Context-Based Captioning is a high-accuracy, live transcription system designed to master complex STEM terminology through a hybrid multi-stage architecture. By integrating course-specific materials via ModernBERT and Shallow Fusion, the pipeline dynamically biases distil-whisper to ensure technical accuracy in real-time university environments.

Research Leads: Christina Xie

Advising Professor: Steve Engels

Context Based Captioning figure 1
Context Based Captioning figure 2
Context Based Captioning figure 3
Context Based Captioning figure 4
Context Based Captioning figure 5
Context Based Captioning figure 6
Context Based Captioning figure 7
AccessibilityExplainable AIReal Time SystemsEd TechSpeech RecognitionASRNLP
GitHubHuggingFace

Machine Unlearning

Evaluating the performance trade-offs of diverse machine unlearning frameworks by simulating user data removal requests within social media recommender systems.

Research Leads: Seoyun Yang

Machine Unlearning figure 1
Machine Unlearning figure 2
Machine Unlearning figure 3
Machine Unlearning figure 4
Machine UnlearningData PrivacyAI SafetyGDPRBenchmark
GitHubProposalSlides

LLM Social Simulation

Studying collective behavior of LLM agents in multi-agent environments with shared resource constraints. Investigates cooperation vs. collapse dynamics, resource competition, social norm emergence, and governance mechanisms when agents compete over regenerating resources.

Research Leads: Clementine Yang

LLM Social Simulation figure 1
LLM Social Simulation figure 2
LLM Social Simulation figure 3
LLM Social Simulation figure 4
Multi-Agent SystemsLLM AgentsSocial SimulationCooperationGame Theory

Have a cool research idea?

Join UsMeet the Team