Hanlin Xue

AAAI Conference 2026 Conference Paper

Detecting Emotional Dynamic Trajectories: An Evaluation Framework for Emotional Support in Language Models

• Zhouxing Tan
• Ruochong Xiong
• Yulong Wan
• Jinlong Ma
• Hanlin Xue
• Qichun Deng
• Haifeng Jing
• Zhengtong Zhang

Emotional support is a core capability in human-AI interaction, with applications including psychological counseling, role play, and companionship. However, existing evaluations of large language models (LLMs) often rely on short, static dialogues and fail to capture the dynamic and long-term nature of emotional support. To overcome this limitation, we shift from snapshot-based evaluation to trajectory-based assessment, adopting a user-centered perspective that evaluates models based on their ability to improve and stabilize user emotional states over time. Our framework constructs a large-scale benchmark consisting of 328 emotional contexts and 1,152 disturbance events, simulating realistic emotional shifts under evolving dialogue scenarios. To encourage psychologically grounded responses, we constrain model outputs using validated emotion regulation strategies such as situation selection and cognitive reappraisal. User emotional trajectories are modeled as a first-order Markov process, and we apply causally-adjusted emotion estimation to obtain unbiased emotional state tracking. Based on this framework, we introduce three trajectory-level metrics: Baseline Emotional Level (BEL), Emotional Trajectory Volatility (ETV), and Emotional Centroid Position (ECP). These metrics collectively capture user emotional dynamics over time and support comprehensive evaluation of long-term emotional support performance of LLMs. Extensive evaluations across a diverse set of LLMs reveal significant disparities in emotional support capabilities and provide actionable insights for model development.

AAAI Conference 2026 Conference Paper

Efficient Transcoder Adaptation for Fine-Tuned Models: Revealing Medical Reasoning Mechanisms in Large Language Models

• Zhouxing Tan
• Hanlin Xue
• Yulong Wan
• Ruochong Xiong
• Xu Chu
• Xiang Li
• Junfei Liu

Large language models (LLMs) suffer from a lack of decision-making transparency, limiting their deployment in high-stakes domains such as healthcare. We propose a mechanistic interpretability framework that introduces two novel paradigms: Medical Fine-Tuning with Frozen Attention Layers (FTFA) and Posterior Adaptation Transcoders (PAT). FTFA freezes attention layers while fine-tuning only feed-forward network (FFN) parameters, enabling PAT to efficiently adapt pre-trained transcoders on the same data. This approach achieves over 1000× efficiency improvement compared to training transcoders from scratch. We theoretically justify this methodology and demonstrate its cost-effectiveness for cross-domain transfer. Transcoders are sparse autoencoders that replace MLP layers to provide interpretable feature representations. By substituting MLP layers of both base Gemma2-2b and its medical fine-tuned variant with per-layer transcoders, we enable feature-level attribution analysis. Through systematic pruning and node merging of resulting attribution graphs, we construct human-interpretable decision pathways. Our analysis reveals that LLMs employ two parallel mechanisms for medical diagnosis: pattern matching and multi-hop reasoning, with fine-tuned models demonstrating enhanced correct reasoning patterns. This work provides a practical framework for training transcoders on fine-tuned models at minimal cost, enabling broader application of mechanistic interpretability across domains and potentially guiding model training through transcoder-based analysis.

JMLR Journal 2015 Journal Article

Supervised Learning via Euler's Elastica Models

• Tong Lin
• Hanlin Xue
• Ling Wang
• Bo Huang
• Hongbin Zha

This paper investigates the Euler's elastica (EE) model for high-dimensional supervised learning problems in a function approximation framework. In 1744 Euler introduced the elastica energy for a 2D curve on modeling torsion-free thin elastic rods. Together with its degenerate form of total variation (TV), Euler's elastica has been successfully applied to low- dimensional data processing such as image denoising and image inpainting in the last two decades. Our motivation is to apply Euler's elastica to high-dimensional supervised learning problems. To this end, a supervised learning problem is modeled as an energy functional minimization under a new geometric regularization scheme, where the energy is composed of a squared loss and an elastica penalty. The elastica penalty aims at regularizing the approximated function by heavily penalizing large gradients and high curvature values on all level curves. We take a computational PDE approach to minimize the energy functional. By using variational principles, the energy minimization problem is transformed into an Euler-Lagrange PDE. However, this PDE is usually high-dimensional and can not be directly handled by common low-dimensional solvers. To circumvent this difficulty, we use radial basis functions (RBF) to approximate the target function, which reduces the optimization problem to finding the linear coefficients of these basis functions. Some theoretical properties of this new model, including the existence and uniqueness of solutions and universal consistency, are analyzed. Extensive experiments have demonstrated the effectiveness of the proposed model for binary classification, multi-class classification, and regression tasks. [abs] [ pdf ][ bib ] &copy JMLR 2015. ( edit, beta )

ICML Conference 2012 Conference Paper

Total Variation and Euler's Elastica for Supervised Learning

• Tong Lin 0002
• Hanlin Xue
• Ling Wang
• Hongbin Zha