I am currently a PhD candidate at Temple University, as a member of Temple AGI Team. My research direction is Artificial General Intelligence. I have been curious about how the mind works. I am dedicated to exploring the principles of intelligence, building a theory of intelligence, creating intelligent machines, and liberating human beings from burdensome mental work.

📖 Education

• 2022.08 – Now, Ph.D. Computer and Information Sciences, Temple University.
• 2019.09 – 2022.06, M.S. Peking University.
• 2015.09 – 2019.06, B. Eng. University of Science and Technology Beijing.

📄 Selected Papers

• Anti-Goal Reasoning: Rethinking the Theory of Goal Reasoning in Non-Axiomatic Logic. Bowen Xu, 2026. [Tech. Report]
  • It provides a minimal extension of Non-Axiomatic Logic to support anti-goal reasoning – i.e., reasoning about avoidance.
• Vision by Logic: a Naïve Visual Recognition Model with Non-Axiomatic Logic. Bowen Xu*, Boyang Xu, and Pei Wang, AGI-Conference 2026. [accepted] | code
  • It demonstrates that visual recognition can be effectively modeled using NAL.
• Hallucination is Inevitable for LLMs with the Open World Assumption. Bowen Xu, 2025. [preprint]
  • The “hallucination” problem of Large Language Models (LLMs) can be reduced to the generalization problem of learners in general.
  • With the open world assumption, generalization error is inevitable.
• On the Essence of Spatial Sense and Objects in Intelligence. 🎖️AGI-25 Best Paper Bowen Xu*, and Pei Wang*, AGI-Conference 2025 | code
  • This paper provides a theoretical foundation for AGI systems to construct spatial sense and objects in the mind.
  • The essence of an object is the sum of the positional relations among its parts, each of which is also an object.
  • The essence of spatial sense (or identically, space) is the sum of equivalence relations among changes in positional representations.
• A Brain-Inspired Sequence Learning Model based on a Logic. Bowen Xu, Scientific Reports, 2025. | code
  • The model is based on Non-axiomatic Logic (NAL) and mimics neocortical mini-columns.
  • The model demonstrates satisfactory performance in online, continual sequence learning, providing preliminary evidence that sequence learning can be effectively and efficiently achieved using NAL.
• What Do You Mean By “Open World”?. Bowen Xu, NeurIPS 2024 Workshop on Open-World Agents.
  • “Open world” is characterized by two properties, unbounded space and time, and unbounded tasks.
• Causal Inference in NARS. Bowen Xu, and Pei Wang*, AGI-Conference 2024, 2024.07
  • Causation is a fat concept. The primary meaning of causation is prediction, while other demands are attached in different contexts. No true causal graph should be assumed.
  • The NARS approach to causal inferene is systematically summarized.
• What is Meant by AGI? On the Definition of Artificial General Intelligence. Bowen Xu, 2024. [preprint]
  • I summarized the genious ideas on the definitions of intelligence proposed in previous works.
  • The definition of AGI is provided, “An Artificial General Intelligence (AGI) system is a computer that is adaptive to the open environment with limited computational resources and that satisfies certain principles.” The terms ‘adaptive’, ‘open environment’, and ‘certain principles’ were specified in the paper.
• Artificial Open World for Evaluating AGI: a Conceptual Design. Bowen Xu*, and Quansheng Ren*, AGI-Conference 2022, 2022.07
  • We argued that the major concern of fairly evaluating AGI in practice is to avoid “the trap of developers’ experience”, which means it is humans rather than machines who solve problems.
  • The measure of intelligence is related to the adaptive (or learning) speed.
• The Gap between Intelligence and Mind. Bowen Xu, Xinyi Zhan, and Quansheng Ren*, AGI-Conference 2021, 2021.10
  • The hard problem of consciousness is discussed in the context of AGI.
  • Feeling is defined as a subjective, private, and non-representable inner experience that manifests as a tendency to adaptively change input by expressing its inner state.

   Blog Posts

• Artificial General Intelligence: What it is, How to Evaluate it, and How to Realize it. Bowen Xu, 2024.
  • [zh-CN] 通用人工智能：是什么？如何测试？如何实现？, 徐博文, 集智俱乐部公众号, 2024.
• [zh-CN] 智能是什么？, 徐博文, 集智俱乐部公众号, 2023.

🎖 Honors&Awards

• 2025, AGI-25 Best Paper Award
• 2019, Outstanding Dissertation Award of Beijing
• 2019, Outstanding Graduate of Beijing
• 2014, China Adolescent Robotics Competition, the National Second Prize
• 2014, Fujian Adolescent Robotics Competition, the Provincial Champion

🤝 Service

• Program Committee, AAAI 2026, 2027.
• Program Committee, AGI Conference, 2023, 2024, 2025, 2026.
• Teaching Assistant for CIS 5511 Programming Techniques, 2025.08 – Now
• Teaching Assistant for CIS 4517/5517 Data-intensive and Cloud Computing, 2024.08 – Now
• Teaching Assistant for CIS 1057 Computer Programming in C, 2022.08 – 2024.05

⚙ Projects

• AGI Survey (in progress)
  • Please contact me if you have suggestions or want to collaborate.
• narsese-vscode-extension: a Narsese (the representation language of NARS) syntax highlighting extension for Visual Studio Code.

© Bowen Xu