Publications

Timing is Everything: Temporal Scaffolding of Semantic Surprise in Humor

Yuxi Ma*, Yongqian Peng*, Junchen Lyu, Chi Zhang, Yixin Zhu

CogSci 2026 (Marr Prize — Best Paper)

A Dual Prediction Violation account of humor: across 828 stand-up performances, temporal scaffolding of semantic surprise — pauses lengthened before high-surprise punchlines — outweighs incongruity alone in predicting laughter.

PDF

Multi-Level Narrative Evaluation Outperforms Lexical Features for Mental Health

Yuxi Ma*, Jieming Cui*, Muyang Li*, Ye Zhao, Yu Li, Yixuan Wang, Chi Zhang, Yinyin Zang, Yixin Zhu

CogSci 2026

A three-tier framework for therapeutic writing: across 830 Chinese texts on depression, anxiety, and trauma, macro-level narrative evaluation — story grammar, coherence, propositional composition — substantially outperforms lexical and embedding features.

PDF

NarrativeLoom: Enhancing Creative Storytelling through Multi-Persona Collaborative Improvisation

Yuxi Ma, Yongqian Peng, Fengyuan Yang, Siyu Zha, Chi Zhang, Zixia Jia, Zilong Zheng, Yixin Zhu

CHI 2026

LLM personas that improvise together, riffing off one another to help people co-write richer, more surprising stories.

PDF Video Project

Learning physics-grounded 4D dynamics with neural gaussian force fields

Shiqian Li, Ruihong Shen, Junfeng Ni, Chang Pan, Chi Zhang, Yixin Zhu

ICLR 2026

Attaches learned force fields to Gaussian particles to predict physically grounded 4D scene dynamics, from collisions to deformation.

PDF Video Project Data Code

Neural Force Field: Few-shot Learning of Generalized Physical Reasoning

Shiqian Li, Ruihong Shen, Yaoyu Tao, Chi Zhang, Yixin Zhu

ICLR 2026

Casts physical reasoning as learning object-wise force fields, generalizing to unseen dynamics from only a handful of examples.

PDF Video Project Data Code

Proposing and solving olympiad geometry with guided tree search

Chi Zhang, Jiajun Song, Siyu Li, Yitao Liang, Yuxi Ma, Wei Wang, Yixin Zhu, Song-Chun Zhu

Nature Machine Intelligence, Jan Issue, 2026

A neuro-symbolic engine that proposes and solves Olympiad-level geometry problems through guided tree search — the January 2026 cover of Nature Machine Intelligence.

In the press 新华社 Xinhua News 人民日报人民网 People's Daily 科技日报

PDF Code

Heterogeneous Adversarial Play in Interactive Environments

Manjie Xu, Xinyi Yang, Jiayu Zhan, Wei Liang, Chi Zhang, Yixin Zhu

NeurIPS 2025

Agents acquire open-ended skills through adversarial play between heterogeneous roles in interactive environments.

PDF Poster Video Project Code

Seed-Prover: Deep and Broad Reasoning for Automated Theorem Proving

ByteDance Seed AI4Math

Tech report, July 2025

Deep-and-broad reasoning for automated theorem proving in Lean — reaching state of the art on competition math and formally reaching silver-medal performance in IMO 2025.

PDF Code

Probing and inducing combinational creativity in vision-language models

Yongqian Peng, Yuxi Ma, Mengmeng Wang, Yuxuan Wang, Yizhou Wang, Chi Zhang, Yixin Zhu, Zilong Zheng

CogSci 2025

Probes whether vision-language models can blend familiar concepts into genuinely novel ones — and shows how to induce that creativity.

In the press 量子位

PDF Video Project Data Code

A simulation-heuristics dual-process model for intuitive physics

Shiqian Li, Yuxi Ma, Jiajun Yan, Bo Dai, Yujia Peng, Chi Zhang, Yixin Zhu

CogSci 2025

Models intuitive physics as a dual process that blends mental simulation with fast heuristics, matching human judgments and timing.

PDF Poster Video Data Code

Human-level few-shot concept induction through minimax entropy learning

Chi Zhang, Baoxiong Jia, Yixin Zhu, Song-Chun Zhu

Science Advances, Vol 10, Issue 16, 2024

Induces abstract visual concepts from only a few examples at human-level accuracy, via a minimax-entropy formulation — published in Science Advances.

In the press 人民日报新华网光明网科技日报

PDF

I-PHYRE: Interactive Physical Reasoning

Shiqian Li, Kewen Wu, Chi Zhang, Yixin Zhu

ICLR 2024

A benchmark for interactive physical reasoning, where agents must act in real time to exploit — not merely predict — physical dynamics.

In the press 北大AI院

PDF Poster Video Project Code

Active Reasoning in an Open-World Environment

Manjie Xu, Guangyuan Jiang, Wei Liang, Chi Zhang, Yixin Zhu

NeurIPS 2023

Agents that actively gather evidence through interaction to answer questions in an open world.

PDF Supp. Poster Project Code

Evaluating and Inducing Personality in Pre-trained Language Models

Guangyuan Jiang*, Manjie Xu*, Song-Chun Zhu, Wenjuan Han, Chi Zhang, Yixin Zhu

NeurIPS 2023 (Spotlight)

Shows pre-trained language models carry measurable, steerable personality traits — with a framework to evaluate and induce them.

PDF Supp. Poster

Interactive Visual Reasoning under Uncertainty

Manjie Xu*, Guangyuan Jiang*, Wei Liang, Chi Zhang, Yixin Zhu

NeurIPS 2023 Datasets and Benchmarks

Agents run experiments to uncover hidden causal 'blicket' rules, benchmarking interactive visual reasoning under uncertainty.

PDF Supp. Poster Project Code

X-VoE: Measuring eXplanatory Violation of Expectation in Physical Events

Bo Dai, Linge Wang, Baoxiong Jia, Zeyu Zhang, Song-Chun Zhu, Chi Zhang, Yixin Zhu

ICCV 2023 (Oral)

Tests whether machines can explain surprising physical events, not just detect them — violation-of-expectation as infants reveal it.

PDF Supp. Poster Project Data Code

MEWL: Few-shot multimodal word learning with referential uncertainty

Guangyuan Jiang, Manjie Xu, Shiji Xin, Wei Liang, Yujia Peng, Chi Zhang, Yixin Zhu

ICML 2023

A few-shot benchmark for learning new words from images under referential uncertainty.

PDF Project Code

Brain in a vat: On missing pieces towards artificial general intelligence in large language models

Yuxi Ma, Chi Zhang, Song-Chun Zhu

Opinion, July 2023

An opinion piece on what today's language models still lack — grounding, embodiment, and agency — on the road to AGI.

PDF

On the Learning Mechanisms in Physical Reasoning

Shiqian Li*, Kewen Wu*, Chi Zhang, Yixin Zhu

NeurIPS 2022 (Spotlight)

Disentangles how models actually acquire physical reasoning, separating the roles of perception, dynamics, and inference.

In the press Institute for AI, PKU

PDF Poster Project Code