Incremental Residual Concept Bottleneck Models

Chenming Shang, Shiji Zhou, Hengyuan Zhang, Xinzhe Ni, Yujiu Yang, Yuwang Wang

[Paper] | [Code]

• Our work primarily addresses the challenges of completeness, purity and precision in previous methods.
• We utilize the multimodal model CLIP to extract and discover concepts, and design interpretable classification model to enhance the interpretability and few-shot generalization ability of the algorithm.
• We propose the Incremental Residual Concept Bottleneck Models (Res-CBM) to address the challenge of concept completeness and the Concept Utilization Efficiency (CUE) metric to measure the descriptive efficiency of CBMs.

Understanding Multimodal Deep Neural Networks: A Concept Selection View

Chenming Shang, Hengyuan Zhang, Hao Wen, Yujiu Yang

[Paper] | [Code]

• We observe the long-tail distribution of concepts, based on which we propose a two-stage Concept Selection Model (CSM) to mine core concepts without introducing any human priors.
• The concept greedy rough selection algorithm is applied to extract head concepts, and then the concept mask fine selection method performs the extraction of core concepts.

Compositional Generalization through Brain-inspired Geometric Constraints on Representation Structure

Chenming Shang, Shiji Zhou, Hengyuan Zhang, Xinchen Zhang, Lei Ke, Yuwang Wang, Yujiu Yang

• We establish a quantitative relationship between Parallelism Score (PS) in cognitive neuroscience and the compositional generalization (CG) ability of model representations.
• To optimize the CG performance theoretically, we propose the incorporation of two regularization techniques, Parallelism Score Maximization and Distance Variance Minimization, to enhance the model CG ability.

HermesFlow: Seamlessly Closing the Gap in Multimodal Understanding and Generation

Ling Yang, Xinchen Zhang, Ye Tian, Chenming Shang, Minghao Xu, Wentao Zhang, Bin Cui

AnyCharV: Bootstrap Controllable Character Video Generation with Fine-to-Coarse Guidance

Zhao Wang, Hao Wen, Lingting Zhu, Chenming Shang, Yujiu Yang, Qi Dou

[Paper] | [Code]

ShifCon: Enhancing Non‑Dominant Language Capabilities with a Shift‑based Contrastive Framework

Hengyuan Zhang*, Chenming Shang*, Sizhe Wang, Dongdong Zhang, Feng Yao, Renliang Sun, Yiyao Yu, Yujiu Yang, Furu Wei (* denotes equal contribution)

[Paper] | [Code]

• Based on the singular value decomposition, we define the language subspace.
• We then transfer the non‑dominant language space to the dominant language space through shift projection and leverage contrastive learning to further alignment.
• The non‑dominant language can access richer information from the parameters of LLMs and improve comprehension capability.

Assisting Language Learners: Automated Trans-Lingual Definition Generation via Contrastive Prompt Learning

Hengyuan Zhang, Dawei Li, Yanran Li, Chenming Shang, Chufan Shi, Yong Jiang

[Paper]

Back-door Intervention based Causal Machine Learning Algorithm for Confounders

Chenming Shang, Zhi Jin

• I complete my bachelor’s thesis under the guidance of Prof. Jin Zhi, and get 95 points, ranking 1st in the major.
• Based on the back-door intervention formula in causal inference, I propose a causal machine learning algorithm for confounders, applying in the recommendation system to solve the problems of popularity bias and conformity bias.

A Question-centric Multi-experts Contrastive Learning Framework for Improving the Accuracy and Interpretability of Deep Sequential Knowledge Tracing Models

Hengyuan Zhang, Zitao Liu, Chenming Shang, Dawei Li, Yong Jiang

[Paper] | [Code]

Seeing Health with Eyes: Feature Combination for Image-Based Human BMI Estimation

Junjia Huang*, Chenming Shang*, Aolin Xiong, Yuxian Pang, Zhi Jin (* denotes equal contribution)

[Paper] | [Code]

Peripheral PD-1+NK Cells Could Predict the 28-day Mortality in Sepsis Patients

Jia Tang, Chenming Shang, Yue Chang, Wei Jiang, Jun Xu, Leidan Zhang, Lianfeng Lu, Ling Chen, Xiaosheng Liu, Qingjia Zeng , Wei Cao, Taisheng Li

[Paper]

Establishment of a Differential Diagnosis Method for AOSD and Lymphoma based on Random Forest Algorithm

Jia Tang*, Chenming Shang*, Yue Chang, Wei Jiang, Jun Xu, Leidan Zhang, Lianfeng Lu, Ling Chen, Xiaosheng Liu, Qingjia Zeng , Wei Cao, Taisheng Li (* denotes equal contribution)