arXiv subjects

Zhao Tan

Publications and source records attributed to Zhao Tan.

10 recordsLinked to original sources

arXiv2026

LEAF-SQL: Level-wise Exploration with Adaptive Fine-graining for Text-to-SQL Skeleton Prediction

Text-to-SQL translates natural language questions into executable SQL queries, enabling intuitive database access for non-experts. While large language models achieve strong performance on Text-to-SQL with prompting, they still struggle with complex queries that involve deeply nested logic or multiple clauses. A widely used approach employs SQL skeletons--intermediate representations of query logic--to streamline generation, but existing methods are limited by their reliance on a single structural hypothesis and lack of progressive reasoning. To overcome these limitations, we propose LEAF-SQL, a novel framework that reframes skeleton prediction as a coarse-to-fine tree search process. LEAF-SQL enables systematic exploration of diverse structural hypotheses with adaptive refinement. Several key techniques are employed in LEAF-SQL: (1) a three-level skeleton hierarchy to guide the search, (2) a Skeleton Formulation Agent to generate diverse candidates, and (3) a Skeleton Evaluation Agent to efficiently prune the search space. This integrated design yields skeleton candidates that are both structurally diverse and granularity-adaptive, providing a stronger foundation for the SQL generation. Extensive experiments show that LEAF-SQL consistently improves the performance of various LLM backbones. On the official hidden test set of the challenging BIRD benchmark, our method achieves 71.6 execution accuracy, which outperforms leading search-based and skeleton-based methods, affirming its effectiveness for complex queries.

cs.CL↗

arXiv2026

Sonar-TS: Search-Then-Verify Natural Language Querying for Time Series Databases

Natural Language Querying for Time Series Databases (NLQ4TSDB) aims to assist non-expert users retrieve meaningful events, intervals, and summaries from massive temporal records. However, existing Text-to-SQL methods are not designed for continuous morphological intents such as shapes or anomalies, while time series models struggle to handle ultra-long histories. To address these challenges, we propose Sonar-TS, a neuro-symbolic framework that tackles NLQ4TSDB via a Search-Then-Verify pipeline. Analogous to active sonar, it utilizes a feature index to ping candidate windows via SQL, followed by generated Python programs to lock on and verify candidates against raw signals. To enable effective evaluation, we introduce NLQTSBench, the first large-scale benchmark designed for NLQ over TSDB-scale histories. Our experiments highlight the unique challenges within this domain and demonstrate that Sonar-TS effectively navigates complex temporal queries where traditional methods fail. This work presents the first systematic study of NLQ4TSDB, offering a general framework and evaluation standard to facilitate future research.

cs.AI↗

arXiv2024

KeyInst: Keyword Instruction for Improving SQL Formulation in Text-to-SQL

Text-to-SQL parsing involves the translation of natural language queries (NLQs) into their corresponding SQL commands. A principal challenge within this domain is the formulation of SQL queries that are not only syntactically correct but also semantically aligned with the natural language input. However, the intrinsic disparity between the NLQ and the SQL poses a significant challenge. In this research, we introduce Keyword Instruction (KeyInst), a novel method designed to enhance SQL formulation by Large Language Models (LLMs). KeyInst essentially provides guidance on pivotal SQL keywords likely to be part of the final query, thus facilitates a smoother SQL query formulation process. We explore two strategies for integrating KeyInst into Text-to-SQL parsing: a pipeline strategy and a single-pass strategy. The former first generates KeyInst for question, which are then used to prompt LLMs. The latter employs a fine-tuned model to concurrently generate KeyInst and SQL in one step. We developed StrucQL, a benchmark specifically designed for the evaluation of SQL formulation. Extensive experiments on StrucQL and other benchmarks demonstrate that KeyInst significantly improves upon the existing Text-to-SQL prompting techniques.

cs.DB↗

arXiv2024

EPI-SQL: Enhancing Text-to-SQL Translation with Error-Prevention Instructions

The conversion of natural language queries into SQL queries, known as Text-to-SQL, is a critical yet challenging task. This paper introduces EPI-SQL, a novel methodological framework leveraging Large Language Models (LLMs) to enhance the performance of Text-to-SQL tasks. EPI-SQL operates through a four-step process. Initially, the method involves gathering instances from the Spider dataset on which LLMs are prone to failure. These instances are then utilized to generate general error-prevention instructions (EPIs). Subsequently, LLMs craft contextualized EPIs tailored to the specific context of the current task. Finally, these context-specific EPIs are incorporated into the prompt used for SQL generation. EPI-SQL is distinguished in that it provides task-specific guidance, enabling the model to circumvent potential errors for the task at hand. Notably, the methodology rivals the performance of advanced few-shot methods despite being a zero-shot approach. An empirical assessment using the Spider benchmark reveals that EPI-SQL achieves an execution accuracy of 85.1\%, underscoring its effectiveness in generating accurate SQL queries through LLMs. The findings indicate a promising direction for future research, i.e. enhancing instructions with task-specific and contextualized rules, for boosting LLMs' performance in NLP tasks.

cs.CL↗

arXiv2023

Divide and Prompt: Chain of Thought Prompting for Text-to-SQL

Chain-of-thought (CoT) prompting combined with large language models (LLMs) have achieved encouraging results on complex reasoning tasks. Text-to-SQL is a critical semantic parsing task that converts natural language questions into SQL statements, involving a complex reasoning process. However, there is little work about using CoT prompting to activate LLM's reasoning capabilities on Text-to-SQL tasks. In this work, we propose a new paradigm for prompting Text-to-SQL tasks, called Divide-and-Prompt, which first divides the task into subtasks, and then approach each subtask through CoT. We present 3 prompting-based methods to enhance the Text-to-SQL ability of LLMs. Experiments show that these prompts guide LLMs to generate Text-to-SQL with higher execution accuracy.

cs.CL↗

arXiv2022

Approximation-free control based on the bioinspired reference model for suspension systems with uncertainty and unknown nonlinearity

Uncertainty and unknown nonlinearity are often inevitable in the suspension systems, which were often solved using fuzzy logic system (FLS) or neural networks (NNs). However, these methods are restricted by the structural complexity of the controller and the huge computing cost. Meanwhile, the estimation error of such approximators is affected by adopted adaptive laws and learning gains. Thus, in view of the above problem, this paper proposes the approximation-free control based on the bioinspired reference model for a class of uncertain suspension systems with unknown nonlinearity. The proposed method integrates the superior vibration suppression of the bioinspired reference model and the structural advantage of the prescribed performance function (PPF) in approximation-free control. Then, the vibration suppression performance is improved, the calculation burden is relieved, and the transient performance is improved, which is analyzed theoretically in this paper. Finally, the simulation results validate the approach, and the comparisons show the advantages of the proposed control method in terms of good vibration suppression, fast convergence, and less calculation burden.

eess.SY↗

arXiv2018

Collaborative Multi-modal deep learning for the personalized product retrieval in Facebook Marketplace

Facebook Marketplace is quickly gaining momentum among consumers as a favored customer-to-customer (C2C) product trading platform. The recommendation system behind it helps to significantly improve the user experience. Building the recommendation system for Facebook Marketplace is challenging for two reasons: 1) Scalability: the number of products in Facebook Marketplace is huge. Tens of thousands of products need to be scored and recommended within a couple hundred milliseconds for millions of users every day; 2) Cold start: the life span of the C2C products is very short and the user activities on the products are sparse. Thus it is difficult to accumulate enough product level signals for recommendation and we are facing a significant cold start issue. In this paper, we propose to address both the scalability and the cold-start issue by building a collaborative multi-modal deep learning based retrieval system where the compact embeddings for the users and the products are trained with the multi-modal content information. This system shows significant improvement over the benchmark in online and off-line experiments: In the online experiment, it increases the number of messages initiated by the buyer to the seller by +26.95%; in the off-line experiment, it improves the prediction accuracy by +9.58%.

cs.IR↗

arXiv2014

Joint Sparse Recovery Method for Compressed Sensing with Structured Dictionary Mismatches

In traditional compressed sensing theory, the dictionary matrix is given a priori, whereas in real applications this matrix suffers from random noise and fluctuations. In this paper we consider a signal model where each column in the dictionary matrix is affected by a structured noise. This formulation is common in direction-of-arrival (DOA) estimation of off-grid targets, encountered in both radar systems and array processing. We propose to use joint sparse signal recovery to solve the compressed sensing problem with structured dictionary mismatches and also give an analytical performance bound on this joint sparse recovery. We show that, under mild conditions, the reconstruction error of the original sparse signal is bounded by both the sparsity and the noise level in the measurement model. Moreover, we implement fast first-order algorithms to speed up the computing process. Numerical examples demonstrate the good performance of the proposed algorithm, and also show that the joint-sparse recovery method yields a better reconstruction result than existing methods. By implementing the joint sparse recovery method, the accuracy and efficiency of DOA estimation are improved in both passive and active sensing cases.

cs.IT↗

arXiv2014

Smoothing and Decomposition for Analysis Sparse Recovery

We consider algorithms and recovery guarantees for the analysis sparse model in which the signal is sparse with respect to a highly coherent frame. We consider the use of a monotone version of the fast iterative shrinkage- thresholding algorithm (MFISTA) to solve the analysis sparse recovery problem. Since the proximal operator in MFISTA does not have a closed-form solution for the analysis model, it cannot be applied directly. Instead, we examine two alternatives based on smoothing and decomposition transformations that relax the original sparse recovery problem, and then implement MFISTA on the relaxed formulation. We refer to these two methods as smoothing-based and decomposition-based MFISTA. We analyze the convergence of both algorithms, and establish that smoothing- based MFISTA converges more rapidly when applied to general nonsmooth optimization problems. We then derive a performance bound on the reconstruction error using these techniques. The bound proves that our methods can recover a signal sparse in a redundant tight frame when the measurement matrix satisfies a properly adapted restricted isometry property. Numerical examples demonstrate the performance of our methods and show that smoothing-based MFISTA converges faster than the decomposition-based alternative in real applications, such as MRI image reconstruction.

math.OC↗

arXiv2013

Direction of Arrival Estimation Using Co-prime Arrays: A Super Resolution Viewpoint

We consider the problem of direction of arrival (DOA) estimation using a newly proposed structure of non-uniform linear arrays, referred to as co-prime arrays, in this paper. By exploiting the second order statistical information of the received signals, co-prime arrays exhibit O(MN) degrees of freedom with only M + N sensors. A sparsity based recovery method is proposed to fully utilize these degrees of freedom. Unlike traditional sparse recovery methods, the proposed method is based on the developing theory of super resolution, which considers a continuous range of possible sources instead of discretizing this range into a discrete grid. With this approach, off-grid effects inherited in traditional sparse recovery can be neglected, thus improving the accuracy of DOA estimation. In this paper we show that in the noiseless case one can theoretically detect up to M N sources with only 2M + N sensors. The noise 2 statistics of co-prime arrays are also analyzed to demonstrate the robustness of the proposed optimization scheme. A source number detection method is presented based on the spectrum reconstructed from the sparse method. By extensive numerical examples, we show the superiority of the proposed method in terms of DOA estimation accuracy, degrees of freedom, and resolution ability compared with previous methods, such as MUSIC with spatial smoothing and the discrete sparse recovery method.

cs.IT↗