My current research focuses on Visual Reasoning and Human-Centered AI, aiming to make human–AI interaction more natural, adaptive, and trustworthy. Before joining the PhD program at UC Berkeley, my research was dedicated to Low-Cost Perception, developing cost-effective visual autonomous systems for the changing world.
Please also see my Google Scholar for the complete publication list.
Visual Reasoning
Developing grounded visual reasoning through verification and correction in vision-language and diffusion models.
Generate, but Verify: Reducing Hallucination in Vision-Language Models with Retrospective Resampling New!
Tsung-Han Wu, Heekyung Lee, Jiaxin Ge, Joseph E Gonzalez, Trevor Darrell, David M. Chan NeurIPS 2025 arxivcodepage
Visual Haystacks: A Vision-Centric Needle-In-A-Haystack Benchmark
Tsung-Han Wu, Giscard Biamby, Jerome Quenum, Ritwik Gupta, Joseph E Gonzalez, Trevor Darrell, David M. Chan ICLR 2025 arxivcodepage
See, Say, and Segment: Teaching LMMs to Overcome False Premises
Tsung-Han Wu*, Giscard Biamby*, David M. Chan, Lisa Dunlap, Ritwik Gupta, Xudong Wang, Joseph E Gonzalez, Trevor Darrell CVPR 2024 arxivcodepage
Self-correcting LLM-controlled Diffusion Models
Tsung-Han Wu*, Long Lian*, Joseph E Gonzalez, Boyi Li, Trevor Darrell CVPR 2024 arxivcodepage
Human-Centered AI
Advancing human--AI interaction under real-world dynamics with a focus on human interruption and preference analyses.
Are Large Reasoning Models Interruptible? New!
Tsung-Han Wu*, Mihran Miroyan*, David M Chan, Trevor Darrell, Narges Norouzi, Joseph E Gonzalez ICML 2026 arxivcodepage
Search Arena: Analyzing Search-Augmented LLMs
Mihran Miroyan*, Tsung-Han Wu*, Logan King, Tianle Li, Jiayi Pan, Xinyan Hu, Wei-Lin Chiang, Anastasios N Angelopoulos, Trevor Darrell, Narges Norouzi, Joseph E Gonzalez ICLR 2026 arxivcodepage
Low-Cost Perception
Reducing perception cost with limited labels (active learning)
D2ADA: Dynamic Density-aware Active Domain Adaptation for Semantic Segmentation
