We present the Adaptive Compact Tree (AC-Tree), a data structure similar to a self-balancing binary search tree (BST) that improves storage and operations for sets of integers with few “holes”. AC-Trees can be imple- mented on top of any classic BST (e.g. AVL trees [3]), holding discontiguous intervals at each node to obtain a compact representation only requiring O(k) storage and supporting efficient O(log k) operations, with k the number of intervals. These properties can improve the performances of applications, like Constraint Program- ming (CP) solvers, that incrementally remove values and intervals on domains initially consisting of one large interval. First experiments show that AC-Trees outperform classic self-balancing BSTs in most cases and im- prove CP solvers performances for problems with large domains.

The motivation of implementing Adaptive Compact Tree is to improve the performance of Constraint Pro- gramming(CP) solvers operations, which is used for en-route conflict resolution in. AC-Trees could not only help CP solver to solve the conflict problem much faster, but also outperform classic BSTs on standard set operations. In addition, we will also present optimizations that improve the performance of the conflict detec- tion for the conflict problem previously mentioned. With these optimizations, the conflict detection phase can be completed in a few seconds. Together with our contribution of the ACTree data structure used in CP solver which accelerates conflict resolution, the total execution time has been reduced significantly, which opens the way to a real-time implementation in an operational context.

[Purpose/Significance] Artificial intelligence, as a key driver of the new wave of technological revolution, has become a focal point in global strategic competition. This study analyzes strategic reports published by U.S. think tanks regarding China’s AI development to explore their cognitive characteristics and developmental trends, providing insights for China to formulate response strategies.[Method/Process] 78 representative re- ports from nine U.S. think tanks were selected as samples. By employing topic modeling methods (LDA and DTM) and textual analysis, the core themes of the reports were categorized and their evolution analyzed. The study systematically examined the U.S. think tanks’ AI strategies regarding China from three dimensions: in- novation drivers, security governance frameworks, and the construction of international discourse power. It also compared the differences in strategic characteristics between the Trump and Biden administrations.[Result/ Conclusion] The study found that U.S. think tanks’ AI strategies concerning China exhibit comprehensiveness and interdisciplinarity, with a focus on innovation development in areas such as education, technology re- search, economic markets, and national security, as well as on security governance frameworks including eth- ics, legal regulations, and social impacts. While think tanks during the Trump administration were guided by Cold War thinking, those under the Biden administration shifted toward technological governance and global rule competition. Based on the cognitive characteristics of U.S. think tanks, it is recommended that China enhance intelligence analysis, strengthen independent technological innovation, and promote international dia- logue and cooperation to safeguard its technological advantages and global discourse power.

With the intensification of population aging, the loneliness problem of the elderly has become a global focus. Loneliness not only seriously affects the physical and mental health of the elderly, but also accelerates the de- cline of their quality of life. As an emerging technology, AI intelligent voice robots have various functions such as health monitoring and interactive entertainment. Western countries have widely applied them to the care and companionship of the elderly and achieved initial results. Based on the CAC model (Cognition - Emotion - Intention Model), this paper takes the subjective cognition of the elderly as the independent variable, the emo- tional response of the elderly as the mediating variable, and uses social surveys as a means. Through empirical testing methods, it verifies that external variable factors such as loneliness trigger the elderly’s internal needs for emotional companionship and health management, and further lead to their different degrees of media use of AI intelligent voice robots. The research findings are as follows: (1) The level of the elderly’s cognition and emotional tendency can positively predict the degree of media use of AI intelligent voice robots by the elderly. (2) The loneliness of the elderly can positively predict the emotional tendency of the elderly. (3) The level of the elderly’s cognition can positively predict the emotional tendency of the elderly, and there is a chain

- mediating relationship between them. (4) The level of the elderly’s cognition and emotional tendency play a complete mediating role in the model. (5) The gender of the elderly plays a moderating role in the above - mentioned chain - mediating relationship. This study attempts to provide an empirical theoretical basis for the effective alleviation of the elderly’s loneliness by AI intelligent voice robots.

Addressing the issues of low annotation efficiency and high cognitive load in the digitization process of an- cient books, this study, grounded in cognitive load theory, proposes an ergonomic solution integrating inter- face optimization and user collaboration mechanisms. By employing an information layering strategy and multi-channel interaction design, a three-level interface architecture—categorized as “primary-auxiliary-sup- plementary”—is constructed, deconstructing the content of ancient books into a core text layer, a folded an- notation layer, and a dynamic floating window layer. Furthermore, by integrating semantic indexing with the BERT model, image restoration with GAN networks, and blockchain certification technology, multimodal con- tent management is optimized. Additionally, a dynamic task allocation model and an intelligent collaboration system are designed, leveraging deep reinforcement learning and Q-Learning algorithms to achieve dynamic matching between user capabilities and task difficulty. Copyright traceability and collaboration efficiency are ensured through the application of Hyperledger Fabric blockchain technology.