Doctoral Thesis Defense of Xufeng Zhao

15 October 2025

On 08.10.2025, our colleague Xufeng Zhao successfully defended his doctoral thesis “Environment Exploration and Autonomous Adaptation in Embodied Agents.” We warmly congratulate him on his graduation and wish him all the best in his next steps! Below is a brief overview of his thesis:

Autonomous agents must not only perceive but also actively explore and adapt. Xufeng’s work develops a conceptual and algorithmic toolkit that (1) uses multimodal intrinsic motivation to steer exploration and representation learning, (2) enables interactive, decision-level perception with LLMs (“chat with the environment”) to reason over rich sensory inputs, (3) strengthens reasoning via logic-guided chain-of-thought, and (4) supports affordance discovery and warm-up policies for faster skill acquisition. Together, these advances move embodied agents toward continual capability expansion beyond task-specific training.

Selected publications from his doctorate:

- Xufeng Zhao, Mengdi Li, Wenhao Lu, Cornelius Weber, Jae Hee Lee, Kun Chu, Stefan Wermter. Enhancing Zero-Shot Chain-of-Thought Reasoning in Large Language Models through Logic . LREC-COLING 2024 (Oral). Torino, Italy, May 2024, pp. 6144–6166.

- Xufeng Zhao, Mengdi Li, Jae Hee Lee, Cornelius Weber, Stefan Wermter. Chat with the Environment: Interactive Multimodal Perception Using Large Language Models . IROS 2023 (Oral), pp. 3590–3596.

- Xufeng Zhao, Cornelius Weber, Muhammad Burhan Hafez, Stefan Wermter. Impact Makes a Sound and Sound Makes an Impact: Sound Guides Representations and Explorations . IROS 2022 (Oral), pp. 2512–2518.

- Mengdi Li*, Xufeng Zhao*, Jae Hee Lee, Cornelius Weber, Stefan Wermter. Internally Rewarded Reinforcement Learning . ICML 2023, PMLR, pp. 20556–20574.

We thank the committee and colleagues for their support and look forward to seeing Xufeng’s future contributions to embodied AI and robotics.