Yuhan Luo

Le Lin, Zihao Zhu, Rainbow Tin Hung Ho et al.

Therapeutic art activities, such as expressive drawing and painting, require the synergy between creative visual production and interactive dialogue. Recent advancements in Multimodal Large Language Models (MLLMs) have expanded the capacity of computing systems to interpret both textual and visual data, offering a new frontier for AI-mediated therapeutic support. This work-in-progress paper introduces an MLLM-powered chatbot that analyzes visual creation in real-time while engaging the creator in reflective conversations. We conducted an evaluation with five experts in art therapy and related fields, which demonstrated the chatbot's potential to facilitate therapeutic engagement, and highlighted several areas for future development, including entryways and risk management, bespoke alignment of user profile and therapeutic style, balancing conversational depth and width, and enriching visual interactivity. These themes provide a design roadmap for designing the future AI-mediated creative expression tools.

Qiya Hu, Yuhan Luo

It is known that the weighted $L^2$ projection operator exhibits approximation properties different from those of the classical $L^2$ projection, in the sense that the $L^2$ error of the weighted $L^2$ projection of an $H^1$ function generally cannot be bounded by the $H^1$ semi-norm of the function. In this paper, we establish sharper $L^2$ error estimates for the weighted $L^2$ projection of an $H^1$ function under general weight distributions. These new estimates show that the $L^2$ errors of the weighted $L^2$ projection can be controlled by the $H^1$ semi-norm of the function, except when the weight distribution is highly irregular, such as those resembling a ``checkerboard" pattern. These results can be applied to more refined analyses of domain decomposition methods and multigrid methods for certain partial differential equations with large jump coefficients.

Jack Chen, Fazhong Liu, Naruto Liu et al.

Large language models (LLMs) excel at mathematical reasoning and logical problem-solving. The current popular training paradigms primarily use supervised fine-tuning (SFT) and reinforcement learning (RL) to enhance the models' reasoning abilities. However, when using SFT or RL alone, there are respective challenges: SFT may suffer from overfitting, while RL is prone to mode collapse. The state-of-the-art methods have proposed hybrid training schemes. However, static switching faces challenges such as poor generalization across different tasks and high dependence on data quality. In response to these challenges, inspired by the curriculum learning-quiz mechanism in human reasoning cultivation, We propose SASR, a step-wise adaptive hybrid training framework that theoretically unifies SFT and RL and dynamically balances the two throughout optimization. SASR uses SFT for initial warm-up to establish basic reasoning skills, and then uses an adaptive dynamic adjustment algorithm based on gradient norm and divergence relative to the original distribution to seamlessly integrate SFT with the online RL method GRPO. By monitoring the training status of LLMs and adjusting the training process in sequence, SASR ensures a smooth transition between training schemes, maintaining core reasoning abilities while exploring different paths. Experimental results demonstrate that SASR outperforms SFT, RL, and static hybrid training methods.

Yuhan Luo, Tao Chen, Decheng Liu

Nowadays, visual data forgery detection plays an increasingly important role in social and economic security with the rapid development of generative models. Existing face forgery detectors still can't achieve satisfactory performance because of poor generalization ability across datasets. The key factor that led to this phenomenon is the lack of suitable metrics: the commonly used cross-dataset AUC metric fails to reveal an important issue where detection scores may shift significantly across data domains. To explicitly evaluate cross-domain score comparability, we propose \textbf{Cross-AUC}, an evaluation metric that can compute AUC across dataset pairs by contrasting real samples from one dataset with fake samples from another (and vice versa). It is interesting to find that evaluating representative detectors under the Cross-AUC metric reveals substantial performance drops, exposing an overlooked robustness problem. Besides, we also propose the novel framework \textbf{S}emantic \textbf{F}ine-grained \textbf{A}lignment and \textbf{M}ixture-of-Experts (\textbf{SFAM}), consisting of a patch-level image-text alignment module that enhances CLIP's sensitivity to manipulation artifacts, and the facial region mixture-of-experts module, which routes features from different facial regions to specialized experts for region-aware forgery analysis. Extensive qualitative and quantitative experiments on the public datasets prove that the proposed method achieves superior performance compared with the state-of-the-art methods with various suitable metrics.