Lihua Liu

Event type discovery in open domains aims to automate the induction of event types from completely unlabeled text data. Conventionally, small models utilize clustering techniques to address this task. Nonetheless, the fully unsupervised nature of these methods results in suboptimal performance by small models in this context. Recently, large language models (LLMs) have demonstrated excellent capabilities in contextual understanding, providing additional relevant information for specific task scenarios, albeit with challenges in precision and cost effectiveness. In this paper, we use LLM to guide the clustering of event texts and distill this process into a fine-tuning task for training smaller pre-trained language models. This approach enables effective event type discovery even in scenarios lacking annotated data. The study unfolds in three stages: in action acquisition, leveraging LLMs to extract type-relevant information from each event text, ensuring that the event representations are particular to task-specific details; in clustering refinement and dual-fine-tune, LLMs refine results from both task-agnostic and task-specific perspectives, with the refinement process designed as fine-tuning tasks under different viewpoints to adjust encoders; and in type generation, post-clustering, LLMs generate meaningful event type labels for each cluster. Experiments show that our method outperforms current state-of-the-art approaches and excels in event type discovery tasks even in completely open-domain with no labeled data.

Causal structure learning plays a crucial role in the current field of artificial intelligence, yet existing causal structure learning methods are susceptible to interference from data sample noise and often become trapped in local optima. To address these challenges, this paper introduces a continuous optimization algorithm based on the curriculum learning framework: CL-NOTEARS. The model utilizes the curriculum loss function during training as a priority evaluation metric for curriculum selection and formulates the sample learning sequence of the model through task-level curricula, thereby enhancing the model’s learning performance. A curriculum-based sample prioritization strategy is employed that dynamically adjusts the training sequence based on variations in loss function values across different samples throughout the training process. The results demonstrate a significant reduction in the impact of sample noise in the data, leading to improved model training performance.

Unmanned Aerial Vehicles (UAVs) have been widely used in localized data collection and information search. However, there are still many practical challenges in real-world operations of UAV search, such as unknown search environments. Specifically, the payoff and cost at each search point are unknown for the planner in advance, which poses a great challenge to decision making. That is, UAV search decisions should be made sequentially in an online manner thereby adapting to the unknown search environment. To this end, this paper initiates the problem of online decision making in UAV search planning, where the drone has limited energy supply as a constraint and has to make an irrevocable decision to search this area or route to the next in an online manner. To overcome the challenge of unknown search environment, a joint-planning approach is proposed, where both route selection and search decision are made in an integrated online manner. The integrated online decision is made through an online linear programming which is proved to be near-optimal, resulting in high information search revenue. Furthermore, this joint-planning approach can be favorably applied to multi-round online UAV search planning scenarios, showing a great superiority in first-mover dominance of gathering information. The effectiveness of the proposed approach is validated in a widely applied dataset, and experimental results show the superior performance of online search decision making.