About Me
I'm Chi Zhang (张驰). I recently joined Peking University as a tenure-track assistant professor, after a year of research at ByteDance Seed and a few wonderful ones at Beijing Institute for General Artificial Intelligence (BIGAI) as a full-time research scientist. Before that, I obtained my Ph.D. degree from Department of Computer Science, University of California - Los Angeles, advised by Professor Song-Chun Zhu.
My recent reseach interests include LLM efficient reasoning, agents and agency, abstract reasoning, active reasoning and applied research for challenging problems.
I used to work with Professor Dit-Yan Yeung at Hong Kong University of Science and Technology and Professor Deng Cai at my alma mater, Zhejiang University.
I serve as
- Conference reviewer: CVPR (from 2019), ICCV (from 2019), ECCV (from 2020), NeurIPS (from 2020), ICLR (from 2020), ICML (from 2021), AAAI (from 2020)
- Journal reviewer: IEEE Transactions on Image Processing (TIP), Psychological Review
Publications
For the latest publications, please refer to my Google Scholar and the lab publication page.
Selected Publications
All Publications
-
Timing is Everything: Temporal Scaffolding of Semantic Surprise in HumorCogSci 2026 Marr Prize Winner
-
Multi-Level Narrative Evaluation Outperforms Lexical Features for Mental HealthCogSci 2026
-
NarrativeLoom: Enhancing Creative Storytelling through Multi-Persona Collaborative ImprovisationCHI 2026
-
Learning physics-grounded 4D dynamics with neural gaussian force fieldsICLR 2026
-
Neural Force Field: Few-shot Learning of Generalized Physical ReasoningICLR 2026
-
Proposing and solving olympiad geometry with guided tree searchNature Machine Intelligence, Jan Issue, 2026
-
Heterogeneous Adversarial Play in Interactive EnvironmentsNeurIPS 2025
-
Seed-Prover: Deep and Broad Reasoning for Automated Theorem ProvingTech report, July, 2025
-
Probing and inducing combinational creativity in vision-language modelsCogSci 2025
-
A simulation-heuristics dual-process model for intuitive physicsCogSci 2025
-
Human-level few-shot concept induction through minimax entropy learningScience Advances, Vol 10, Issue 16, 2024
-
I-PHYRE: Interactive Physical ReasoningICLR 2024
-
Active Reasoning in an Open-World EnvironmentNeurIPS 2023
-
Evaluating and Inducing Personality in Pre-trained Language ModelsNeurIPS 2023Spotlight
-
Interactive Visual Reasoning under UncertaintyNeurIPS 2023 Datasets and Benchmarks
-
X-VoE: Measuring eXplanatory Violation of Expectation in Physical EventsICCV 2023Oral
-
MEWL: Few-shot multimodal word learning with referential uncertaintyICML 2023
-
Brain in a vat: On missing pieces towards artificial general intelligence in large language modelsOpinion, July, 2023
-
On the Learning Mechanisms in Physical ReasoningNeurIPS 2022Spotlight
-
Learning Algebraic Representation for Systematic Generalization in Abstract ReasoningECCV 2022Paper / Supp / Poster / Code / Dataset / Project / Media: Institute for AI, PKU / Media: PKU News / Media: PKU Weibo
-
DETR++: Taming Your Multi-Scale Detection TransformerT4V: Transformers for Vision @ CVPR 2022.
-
ACRE: Abstract Causal REasoning Beyond CovariationCVPR 2021
-
Abstract Spatial-Temporal Reasoning via Probabilistic Abduction and ExecutionCVPR 2021
-
Congestion-aware Multi-agent Trajectory Prediction for Collision AvoidanceICRA 2021
-
Dark, Beyond Deep: A Paradigm Shift to Cognitive AI with Humanlike Common SenseEngineering, Volume 6, Issue 3
-
Machine Number Sense: A Dataset of Visual Arithmetic Problems for Abstract and Relational ReasoningAAAI 2020Oral
-
Learning Perceptual Inference by ContrastingNeurIPS 2019Spotlight (2.43% acceptance rate)
-
Learning Virtual Grasp with Failed Demonstrations via Bayesian Inverse Reinforcement LearningIROS 2019
-
RAVEN: A Dataset for Relational and Analogical Visual rEasoNingCVPR 2019
-
MetaStyle: Three-Way Trade-Off Among Speed, Flexibility, and Quality in Neural Style TransferAAAI 2019
-
Mirroring without Overimitation: Learning Functionally Equivalent Manipulation ActionsAAAI 2019Paper /
-
Learning Unmanned Aerial Vehicle Control for Autonomous Target FollowingIJCAI 2018Paper /
-
Question Retrieval for Community-based Question Answering via Heterogeneous Social Influential NetworkNeurocomputing, Volume 285Paper /
-
A Method of Exact 3D Modeling Based on Natural Gestures via Data Gloves (in Chinese)Patent publication number: CN104778746 B
Experience
Tenure-Track Assistant Professor
Peking University
Beijing, China
Research Scientist
ByteDance Seed
Beijing, China
Research Scientist
Beijing Institute for General Artificial Intelligence (BIGAI)
Beijing, China
Education
Ph.D. in Computer Science
University of California, Los Angeles
Los Angeles, California, USA
M.S. in Computer Science
University of California, Los Angeles
Los Angeles, California, USA
B.Eng. in Computer Science
Zhejiang University
Hangzhou, Zhejiang, China
Friends
People I have had the privilege to learn and work with.