Large language models (LLMs) have exhibited remarkable capabilities and achieved significant breakthroughs across various domains, leading to their widespread adoption in recent years. Building on this progress, we investigate their potential in the realm of local life services. In this study, we establish a comprehensive benchmark and systematically evaluate the performance of diverse LLMs across a wide range of tasks relevant to local life services. To further enhance their effectiveness, we explore two key approaches: model fine-tuning and agent-based workflows. Our findings reveal that even a relatively compact 7B model can attain performance levels comparable to a much larger 72B model, effectively balancing inference cost and model capability. This optimization greatly enhances the feasibility and efficiency of deploying LLMs in real-world online services, making them more practical and accessible for local life applications. Available resources are at https://github.com/tsinghua-fib-lab/LocalEval.

Recommender systems that learn from implicit feedback often use large volumes of a single type of implicit user feedback, such as clicks, to enhance the prediction of sparse target behavior such as purchases. Using multiple types of implicit user feedback for such target behavior prediction purposes is still an open question. Existing studies that attempted to learn from multiple types of user behavior often fail to: (i) learn universal and accurate user preferences from different behavioral data distributions, and (ii) overcome the noise and bias in observed implicit user feedback.

Learning reinforcement learning (RL)-based recommenders from historical user-item interaction sequences is vital to generate high-reward recommendations and improve long-term cumulative benefits. However, existing RL recommendation methods encounter difficulties (i) to estimate the value functions for states which are not contained in the offline training data, and (ii) to learn effective state representations from user implicit feedback due to the lack of contrastive signals.

Session-based recommendation aims to predict user behaviors based on anonymous sessions. Recently, session sequences are modeled as graph-structured data. Based on the session graphs, Graph Neural Networks (GNNs) can capture complex transitions of items, compared with previous conventional sequential methods. However, the existing graph-construction approaches have limited power in capturing the position information of items in the session sequences. In addition, GNNs employed in the existing session-based recommendation are not capable to attend over their neighborhoods' features in feature aggregation phase. In this paper, we propose a Position-Aware Gated Graph Attention Network (PAGGAN). Specifically, a reverse-position mechanism is proposed to assign position embeddings to nodes in the session graphs based on the order of items in each session sequence. And we enhance Gated Graph Neural Network (GGNN) by introducing self-attention mechanism when aggregating features from nodes. Experimental results on two real-world datasets show that the PA-GGAN outperforms state-of-the-art methods.