Zhi Zhou (周植)
About
I am a final-year Ph.D. student at the Department of Computer Science and Technology of Nanjing University, advised by Professor Yu-Feng Li (李宇峰), and a member of the LAMDA Group led by Professor Zhi-Hua Zhou (周志华).
| 2022 – Present | Ph.D. in Computer Science, Nanjing University (transferred from the M.Sc. program after the second year) |
| 2020 – 2022 | M.Sc. study in Computer Science, Nanjing University |
| 2016 – 2020 | B.Sc. in Computer Science, Tang Aoqing Honors Program, Jilin University |
Research
Our long-term goal is to advance Lifelong Intelligence toward Artificial General Intelligence (AGI). We study how intelligent systems can continuously acquire and refine capabilities by learning from data, interaction, and knowledge through reasoning, with a focus on adaptive learning, agent learning, and neuro-symbolic learning. Our research targets open and dynamic applications, including digital agents, embodied agents, finance, and industry.
Adaptive learning
Studies continual adaptation under distribution shift, task evolution, and non-stationary environments.
Agent learning
Studies agents that acquire policies and skills through sequential interaction and feedback.
Neuro-symbolic learning
Combines neural representations with symbolic structure to support compositional reasoning and reliable generalization.
Recent News
- 2026.09Three papers accepted by EMNLP 2026 — phsical and web agents, symbolic feature engineering.
- 2026.07One paper accepted by ECCV 2026 — a neuro-symbolic benchmark for constrained route planning in remote sensing.
- 2026.05Three papers accepted by ICML 2026 — visual-tabular learning, semi-supervised learning, and test-time adaptation learnability.
- 2026.01One paper accepted by ICLR 2026 — a formal subgoal-completion benchmark for machine learning theory.
- 2025.11One paper accepted by AAAI 2026 — an LLM reasoning paradigm.
- 2025.09One paper accepted by NeurIPS 2025 — a theoretical analysis of LLM reasoning.
- 2025.08One paper accepted by EMNLP 2025 for oral presentation — test-time scaling of LLM reasoning.
- 2025.05Two papers accepted by ICML 2025 — vision-language model selection & reuse, and a tabular feature-shift benchmark.
- 2025.04Four papers accepted by IJCAI 2025 — generative model identification, a neuro-symbolic reasoning survey, tabular test-time adaptation, and prompt learning for VLMs.
- 2025.01One paper accepted by ICLR 2025 — neural theorem proving with diversified tactic calibration.
- 2025.01Awarded a research grant from the NSFC Research Program for Young Students — Robust Weakly-Supervised Learning for Open Environment (624B2068).
- 2024.12One paper accepted by AAAI 2025 — fully test-time adaptation for tabular data.
- 2024.09One paper accepted by NeurIPS 2024 — neuro-symbolic data generation for math reasoning.
- 2024.05Two papers accepted by ICML 2024 — robust prompt tuning with OOD detection, and long-tail learning with foundation models.
- 2024.04One paper accepted by IJCAI 2024 — spectrally localized augmentation for graph consistency learning.
- 2024.01One paper accepted by ICLR 2024 — realistic evaluation of semi-supervised learning in open environments.
- 2023.12Two papers accepted by AAAI 2024 — robust test-time adaptation and high-order graph attention networks.
- 2023.04Three papers (two Oral) accepted by ICML 2023 — open-world test-time adaptation, bidirectional semi-supervised learning, and identifying useful learnwares.
Selected Publications
full list →
* equal contribution · ✉ corresponding author
