Chunlin Gong

Zhao Tong, Chunlin Gong, Yiping Zhang et al.

From generating headlines to fabricating news, the Large Language Models (LLMs) are typically assessed by their final outputs, under the safety assumption that a refusal response signifies safe reasoning throughout the entire process. Challenging this assumption, our study reveals that during fake news generation, even when a model rejects a harmful request, its Chain-of-Thought (CoT) reasoning may still internally contain and propagate unsafe narratives. To analyze this phenomenon, we introduce a unified safety-analysis framework that systematically deconstructs CoT generation across model layers and evaluates the role of individual attention heads through Jacobian-based spectral metrics. Within this framework, we introduce three interpretable measures: stability, geometry, and energy to quantify how specific attention heads respond or embed deceptive reasoning patterns. Extensive experiments on multiple reasoning-oriented LLMs show that the generation risk rise significantly when the thinking mode is activated, where the critical routing decisions concentrated in only a few contiguous mid-depth layers. By precisely identifying the attention heads responsible for this divergence, our work challenges the assumption that refusal implies safety and provides a new understanding perspective for mitigating latent reasoning risks.

Chunlin Gong, Xu Li, Hua Su et al.

In this paper, a new sequential surrogate-based optimization (SSBO) algorithm is developed, which aims to improve the global search ability and local search efficiency for the global optimization of expensive black-box models. The proposed method involves three basic sub-criteria to infill new samples asynchronously to balance the global exploration and local exploitation. First, to capture the promising possible global optimal region, searching for the global optimum with genetic algorithm (GA) based on the current surrogate models of the objective and constraint functions. Second, to infill samples in the region with sparse samples to improve the global accuracy of the surrogate models, a grid searching with Latin hypercube sampling (LHS) with the current surrogate model is adopted to explore the sample space. Third, to accelerate the local searching efficiency, searching for a local optimum with sequential quadratic programming (SQP) based on the local surrogate models in the reduced interval, which involves some samples near the current optimum. When the new sample is too close to the existing ones, the new sample should be abandoned, due to the poor additional information. According to the three sub-criteria, the new samples are placed in the regions which have not been fully explored and includes the possible global optimum point. When a possible global optimum point is found, the local searching sub-criterion captures the local optimum around it rapidly. Numerical and engineering examples are used to verify the efficiency of the proposed method. The statistical results show that the proposed method has good global searching ability and efficiency.