I’m a CS Ph.D. student in the NLP Lab at UC Santa Barbara (UCSB), advised by Prof. Shiyu Chang and Dr. Yang Zhang (MIT-IBM Watson AI Lab). I’m currently a research intern at 
Salesforce AI Research, hosted by Dr. Semih Yavuz and Prof. Shafiq Joty. I have also closely collaborated with Prof. Wenhu Chen at TIGER Lab.
Previously, I earned my bachelor’s degree from Nanjing University and my master’s degree from Johns Hopkins University. Before starting my Ph.D., I was a full-time applied scientist at 
Amazon Alexa AI / AGI.
My current research interests are self-evolving agents, agentic RL, and multimodal learning. I have previously worked on conversational AI, information retrieval, and LLMs for personalization.
Education
-
UC Santa Barbara Ph.D. in Computer Science -
Johns Hopkins University M.S. in Electrical and Computer Engineering
-
Nanjing University B.S. in Acoustics & Physics
Industry Experience
-
Salesforce AI Research Research Intern -
Amazon Applied Scientist II (Alexa AI & AGI)
News
-
We released WorldAuditBench, a benchmark for interactive 3D world auditing with multimodal agents.
-
MMEB-V3 has been accepted to COLM 2026.
-
I will be joining Salesforce AI Research as a research intern this coming summer.
-
VLM2Vec-V2 has been accepted to TMLR.
-
I’ve moved my personal homepage to this new site; the old one is now deprecated.
Publications * Equal contribution / co-first author
Full publications on Google Scholar. Citations
-
WorldAuditBench: Interactive 3D World Auditing with Multimodal Agents
AgentsMultimodal LearningEvaluation & Benchmark -
VisualSkill: Multimodal Skills for Computer-Use Agents
AgentsMultimodal Learning -
MMEB-V3: Measuring the Performance Gaps of Omni-Modality Embedding Models
Multimodal LearningEvaluation & Benchmark -
Universal Retrieval for Multimodal Trajectory Modeling
AgentsRetrieval & RAGMultimodal LearningEvaluation & Benchmark -
VLM2Vec-V2: Advancing Multimodal Embedding for Videos, Images, and Visual Documents
Multimodal LearningRetrieval & RAGEvaluation & Benchmark -
StructEval: Benchmarking LLMs’ Capabilities to Generate Structural Outputs
Evaluation & Benchmark -
VLM2Vec: Training Vision-Language Models for Massive Multimodal Embedding Tasks
Multimodal LearningRetrieval & RAGEvaluation & Benchmark -
MEGA-Bench: Scaling Multimodal Evaluation to over 500 Real-World Tasks
Multimodal LearningEvaluation & Benchmark -
Knowledge Enhanced Multi-Domain Recommendations in an AI Assistant Application
Personalization -
LongRAG: Enhancing Retrieval-Augmented Generation with Long-context LLMs
Retrieval & RAG -
MMLU-Pro: A More Robust and Challenging Multi-Task Language Understanding Benchmark
Evaluation & Benchmark -
VideoScore: Building Automatic Metrics to Simulate Fine-grained Human Feedback for Video Generation
Multimodal LearningEvaluation & Benchmark -
Semi-Supervised Reward Modeling via Iterative Self-Training
Post-training -
Position-Aware Parameter Efficient Fine-Tuning Approach for Reducing Positional Bias in LLMs
Post-training -
RecMind: Large Language Model Powered Agent For Recommendation
AgentsPersonalization -
Graph Meets LLM: A Novel Approach to Collaborative Filtering for Robust Conversational Understanding
Conversational AIPersonalization -
PALR: Personalization Aware LLMs for Recommendation
Personalization -
KG-ECO: Knowledge Graph Enhanced Entity Correction for Query Rewriting
Conversational AI -
PENTATRON: PErsonalized coNText-Aware Transformer for Retrieval-based cOnversational uNderstanding
Conversational AIPersonalizationRetrieval & RAG -
Personalized search-based query rewrite system for conversational ai
Conversational AIPersonalizationRetrieval & RAG -
#HowYouTagTweets: Learning User Hashtagging Preferences via Personalized Topic Attention
Personalization -
Black-box Attacks on Spoofing Countermeasures Using Transferability of Adversarial Examples
Conversational AI -
A Comparative Study on Transformer vs RNN in Speech Applications
Conversational AI
Services
- Conference Reviewer: NeurIPS (2024–), ICLR (2024–), ICML (2026–), ACL ARR (2023–), ICASSP (2024–)
- Journal Reviewer: IEEE Access, TOIS, TMLR
- Workshop Organizer: The First Workshop on Personalized Generative AI (@CIKM2023), The Second Workshop on Generative Information Retrieval (@SIGIR2024)