Jinzhou (Marcus) Tang

Jinzhou (Marcus) Tang

MSc in Computer Science

Posts by Collection

publications

VertFound: Synergizing Semantic and Spatial Understanding for Fine-Grained Vertebrae Classification via Foundation Models

Published in MICCAI 2024, 2024

A framework that merges semantic information from CLIP and spatial features from SAM for fine-grained vertebrae classification, with a vertebral positioning with cross-model synergy module and Wasserstein loss.

Recommended citation: Y. Wu, J. Tang, Z. Yao, M. Li, Y. Hong, D. Yu, Z. Gao, B. Chen, and S. Zhao. (2024). "VertFound: Synergizing Semantic and Spatial Understanding for Fine-Grained Vertebrae Classification via Foundation Models." Medical Image Computing and Computer Assisted Intervention (MICCAI), pp. 763–772.
Download Paper

HiVA: Self-Organized Hierarchical Variable Agent via Goal-Driven Semantic-Topological Evolution

Published in AAAI 2025, 2025

A multi-agent framework that models agentic workflows as self-organized graphs and co-evolves agent semantics and collaboration topology via the Semantic-Topological Evolution (STEV) algorithm.

Recommended citation: J. Tang, J. Zhang, Q. Lv, S. Liu, J. Yang, C. Tang, and K. Wang. (2025). "HiVA: Self-Organized Hierarchical Variable Agent via Goal-Driven Semantic-Topological Evolution." Proceedings of the AAAI Conference on Artificial Intelligence (AAAI).
Download Paper

DreamSAC: Learning Hamiltonian World Models via Symmetry Exploration

Published in CVPR 2026, 2026

An RL framework that learns physics-grounded Hamiltonian world models through symmetry exploration, using Hamiltonian-based curiosity to collect physically informative data for extrapolative generalization.

Recommended citation: J. Tang, F. Feng, M. Fu, W. Lin, B. Huang, and K. Wang. (2026). "DreamSAC: Learning Hamiltonian World Models via Symmetry Exploration." IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR).
Download Paper

Self-Improving World Models via Asymmetric Forward-Inverse Consistency

Published in ICLR 2026 Workshop on Recursive Self-Improvement (Spotlight), 2026

A self-improving framework (World Action Verifier) that decomposes action-conditioned prediction into state plausibility and action reachability, exploiting forward–inverse asymmetry for verification and data-efficient world-model learning.

Recommended citation: Y. Liu, L. Kong, F. Feng, W. Lu, J. Tang, X. Zhang, K. Zhang, K. Murphy, Y. Du, and C. Finn. (2026). "Self-Improving World Models via Asymmetric Forward-Inverse Consistency." ICLR 2026 Workshop on Recursive Self-Improvement (RSI), Spotlight.
Download Paper