Text-to-SQL is a fundamental yet challenging task in the NLP area, aiming at translating natural language questions into SQL queries. While recent advances in large language models have greatly improved performance, most existing approaches depend on models with tens of billions of parameters or costly APIs, limiting their applicability in resource-constrained environments. For real world, especially on edge devices, it is crucial for Text-to-SQL to ensure cost-effectiveness. Therefore, enabling the light-weight models for Text-to-SQL is of great practical significance. However, smaller LLMs often struggle with complicated user instruction, redundant schema linking or syntax correctness. To address these challenges, we propose MCTS-SQL, a novel framework that uses Monte Carlo Tree Search to guide SQL generation through multi-step refinement. Since the light-weight models' weak performance of single-shot prediction, we generate better results through several trials with feedback. However, directly applying MCTS-based methods inevitably leads to significant time and computational overhead. Driven by this issue, we propose a token-level prefix-cache mechanism that stores prior information during iterations, effectively improved the execution speed. Experiments results on the SPIDER and BIRD benchmarks demonstrate the effectiveness of our approach. Using a small open-source Qwen2.5-Coder-1.5B, our method outperforms ChatGPT-3.5. When leveraging a more powerful model Gemini 2.5 to explore the performance upper bound, we achieved results competitive with the SOTA. Our findings demonstrate that even small models can be effectively deployed in practical Text-to-SQL systems with the right strategy.

Face detection is a fundamental and practical problem in computer vision, which aims to indicate the face positions in a wild environment precisely. However, different from the generic object detection tasks, there are a large number of face samples that suffer from unconstrained poses, occlusion, extreame lights, or other harmful conditions. YOLOV5 is an incredible milestone in the object detection area, but still not powerful enough for the challenging samples. To ease these difficulties, in this paper, we customize DSH-YOLOV5, a practical face detector. Specifically, we integrate a decoupled head with swin transformer layers, whose self-attention mechanism has great potential to explore subtle interconnection details. Additionally, we use two context modules (CBAM and SSH) to enhance the performance of features. Furthermore, we design a novel copy-paste data augmentation to fit the above challenging scenes. Extensive experiments demonstrate that we achieve SOTA performance on WIDER FACE, FDDB, and PASCAL FACE with competitive computational costs. Moreover, under COVID-19, we add the breathing mask and gender classification branches based on the ROI Align to produce more practical face information.