Danny Dongning Sun
Research Subjects
Danny Dongning Sun has a total of 35 co-authors. Here are the top co-authors by number of shared publications:
- Xiao-Ping Zhang
Tsinghua University
3 shared publications - Bo Li
Northwestern Polytechnical University
2 shared publications - Hongkang Zhang
Tsinghua Shenzhen International Graduate School, Tsinghua University, Shenzhen, China
2 shared publications - Junzhe Jiang
The Hong Kong Polytechnic University
2 shared publications - Shao-Lun Huang
Tsinghua University
2 shared publications - Yang Liu
Department of Obstetrics and Gynecology, The Second Affiliated Hospital of Kunming Medical University, China
2 shared publications
Latest Publications
- 26 September 2026
- Published by Springer Nature
- p. 495-507
Assessing Uncertainty in Stock Returns: A Gaussian Mixture Distribution-Based Method
- 26 September 2026
- Published by Springer Nature
- p. 495-507
Assessing Uncertainty in Stock Returns: A Gaussian Mixture Distribution-Based Method
- Open Access
- 8 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 12327-12338
FactorMiner: A Self-Evolving Agent with Skills and Experience Memory for Financial Alpha Discovery
- 8 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 12327-12338
Open AccessFactorMiner: A Self-Evolving Agent with Skills and Experience Memory for Financial Alpha Discovery
- 4 August 2026
- Published by Springer Nature
- p. 468-482
PSformer: Parameter-Efficient Transformer with Segment Shared Attention for Time Series Forecasting
- 4 August 2026
- Published by Springer Nature
- p. 468-482
PSformer: Parameter-Efficient Transformer with Segment Shared Attention for Time Series Forecasting
- 2 April 2026
- Published by Elsevier in Neural Networks
- Vol. 201, 108924
Towards robust deep reinforcement learning-based quantitative trading with neuro-symbolic trend analysis
- 2 April 2026
- Published by Elsevier in Neural Networks
- Vol. 201, 108924
Towards robust deep reinforcement learning-based quantitative trading with neuro-symbolic trend analysis
- Open Access
Investigating Training Data Detection in AI Coders
Recent advances in code large language models (CodeLLMs) have made them indispensable tools in modern software engineering. However, these models occasionally produce outputs that contain proprietary or sensitive code snippets, raising concerns about potential non-compliant use of training data, and posing risks to privacy and intellectual property. To ensure responsible and compliant deployment of CodeLLMs, training data detection (TDD) has become a critical task. While recent TDD methods have shown promise in natural language settings, their effectiveness on code data remains largely underexplored. This gap is particularly important given code's structured syntax and distinct similarity criteria compared to natural language. To address this, we conduct a comprehensive empirical study of seven state-of-the-art TDD methods on source code data, evaluating their performance across eight CodeLLMs. To support this evaluation, we introduce CodeSnitch, a function-level benchmark dataset comprising 9,000 code samples in three programming languages, each explicitly labeled as either included or excluded from CodeLLM training. Beyond evaluation on the original CodeSnitch, we design targeted mutation strategies to test the robustness of TDD methods under three distinct settings. These mutation strategies are grounded in the well-established Type-1 to Type-4 code clone detection taxonomy. Our study provides a systematic assessment of current TDD techniques for code and offers insights to guide the development of more effective and robust detection methods in the future.
Open AccessInvestigating Training Data Detection in AI Coders
Recent advances in code large language models (CodeLLMs) have made them indispensable tools in modern software engineering. However, these models occasionally produce outputs that contain proprietary or sensitive code snippets, raising concerns about potential non-compliant use of training data, and posing risks to privacy and intellectual property. To ensure responsible and compliant deployment of CodeLLMs, training data detection (TDD) has become a critical task. While recent TDD methods have shown promise in natural language settings, their effectiveness on code data remains largely underexplored. This gap is particularly important given code's structured syntax and distinct similarity criteria compared to natural language. To address this, we conduct a comprehensive empirical study of seven state-of-the-art TDD methods on source code data, evaluating their performance across eight CodeLLMs. To support this evaluation, we introduce CodeSnitch, a function-level benchmark dataset comprising 9,000 code samples in three programming languages, each explicitly labeled as either included or excluded from CodeLLM training. Beyond evaluation on the original CodeSnitch, we design targeted mutation strategies to test the robustness of TDD methods under three distinct settings. These mutation strategies are grounded in the well-established Type-1 to Type-4 code clone detection taxonomy. Our study provides a systematic assessment of current TDD techniques for code and offers insights to guide the development of more effective and robust detection methods in the future.
- Open Access
- 11 April 2025
- Published by Association for the Advancement of Artificial Intelligence (AAAI) in Proceedings of the AAAI Conference on Artificial Intelligence
- Vol. 39 (17) , 18584-18592
Logic-Q: Improving Deep Reinforcement Learning-based Quantitative Trading via Program Sketch-based Tuning
Deep reinforcement learning (DRL) has revolutionized quantitative trading (Q-trading) by achieving decent performance without significant human expert knowledge. Despite its achievements, we observe that the current state-of-the-art DRL models are still ineffective in identifying the market trends, causing them to miss good trading opportunity or suffer from large drawdowns when encountering market crashes. To address this limitation, a natural approach is to incorporate human expert knowledge in identifying market trends. Whereas, such knowledge is abstract and hard to be quantified. In order to effectively leverage abstract human expert knowledge, in this paper, we propose a universal logic-guided deep reinforcement learning framework for Q-trading, called Logic-Q. In particular, Logic-Q adopts the program synthesis by sketching paradigm and introduces a logic-guided model design that leverages a lightweight, plug-and-play market trend-aware program sketch to determine the market trend and correspondingly adjusts the DRL policy in a post-hoc manner. Extensive evaluations of two popular quantitative trading tasks demonstrate that Logic-Q can significantly improve the performance of previous state-of-the-art DRL trading strategies.
- 11 April 2025
- Published by Association for the Advancement of Artificial Intelligence (AAAI) in Proceedings of the AAAI Conference on Artificial Intelligence
- Vol. 39 (17) , 18584-18592
Open AccessLogic-Q: Improving Deep Reinforcement Learning-based Quantitative Trading via Program Sketch-based Tuning
Deep reinforcement learning (DRL) has revolutionized quantitative trading (Q-trading) by achieving decent performance without significant human expert knowledge. Despite its achievements, we observe that the current state-of-the-art DRL models are still ineffective in identifying the market trends, causing them to miss good trading opportunity or suffer from large drawdowns when encountering market crashes. To address this limitation, a natural approach is to incorporate human expert knowledge in identifying market trends. Whereas, such knowledge is abstract and hard to be quantified. In order to effectively leverage abstract human expert knowledge, in this paper, we propose a universal logic-guided deep reinforcement learning framework for Q-trading, called Logic-Q. In particular, Logic-Q adopts the program synthesis by sketching paradigm and introduces a logic-guided model design that leverages a lightweight, plug-and-play market trend-aware program sketch to determine the market trend and correspondingly adjusts the DRL policy in a post-hoc manner. Extensive evaluations of two popular quantitative trading tasks demonstrate that Logic-Q can significantly improve the performance of previous state-of-the-art DRL trading strategies.
- 18 June 2023
- Published by With Intelligence LLC in The Journal of Financial Data Science
Kernel Market Impact Analysis in China A-Share Markets
With transaction-level market data for stocks in China A-share markets, the authors construct individual stocks’ kernel functions of market impact and analyze their statistical properties. Attribution analysis of such kernel functions is also performed to understand how market microstructure variables such as bid–ask spread and liquidity distribution in order books can be used to classify different groups of kernel functions. The authors’ analysis shows that stocks in China A-share markets exhibit clear patterns of market impact curves, which is likely due to specific market structure regulations such as constant tick size across different stocks and stock-specific order book dynamics resulting from market participants’ behaviors. The authors also explore the application of kernel functions in forecasting price movement in close-to-reality trading simulators that consider market impact costs at individual trade level.
- 18 June 2023
- Published by With Intelligence LLC in The Journal of Financial Data Science
Kernel Market Impact Analysis in China A-Share Markets
With transaction-level market data for stocks in China A-share markets, the authors construct individual stocks’ kernel functions of market impact and analyze their statistical properties. Attribution analysis of such kernel functions is also performed to understand how market microstructure variables such as bid–ask spread and liquidity distribution in order books can be used to classify different groups of kernel functions. The authors’ analysis shows that stocks in China A-share markets exhibit clear patterns of market impact curves, which is likely due to specific market structure regulations such as constant tick size across different stocks and stock-specific order book dynamics resulting from market participants’ behaviors. The authors also explore the application of kernel functions in forecasting price movement in close-to-reality trading simulators that consider market impact costs at individual trade level.
- Open Access
Optimal Market Making in the Chinese Stock Market: A Stochastic Control and Scenario Analysis
Market making plays a crucial role in providing liquidity and maintaining stability in financial markets, making it an essential component of well-functioning capital markets. Despite its importance, there is limited research on market making in the Chinese stock market, which is one of the largest and most rapidly growing markets globally. To address this gap, we employ an optimal market making framework with an exponential CARA-type (Constant Absolute Risk Aversion) utility function that accounts for various market conditions, such as price drift, volatility, and stamp duty, and is capable of describing 3 major risks (i.e., inventory, execution and adverse selection risks) in market making practice, and provide an in-depth quantitative and scenario analysis of market making in the Chinese stock market. Our numerical experiments explore the impact of volatility on the market maker's inventory. Furthermore, we find that the stamp duty rate is a critical factor in market making, with a negative impact on both the profit of the market maker and the liquidity of the market. Additionally, our analysis emphasizes the significance of accurately estimating stock drift for managing inventory and adverse selection risks effectively and enhancing profit for the market maker. These findings offer valuable insights for both market makers and policymakers in the Chinese stock market and provide directions for further research in designing effective market making strategies and policies.
Open AccessOptimal Market Making in the Chinese Stock Market: A Stochastic Control and Scenario Analysis
Market making plays a crucial role in providing liquidity and maintaining stability in financial markets, making it an essential component of well-functioning capital markets. Despite its importance, there is limited research on market making in the Chinese stock market, which is one of the largest and most rapidly growing markets globally. To address this gap, we employ an optimal market making framework with an exponential CARA-type (Constant Absolute Risk Aversion) utility function that accounts for various market conditions, such as price drift, volatility, and stamp duty, and is capable of describing 3 major risks (i.e., inventory, execution and adverse selection risks) in market making practice, and provide an in-depth quantitative and scenario analysis of market making in the Chinese stock market. Our numerical experiments explore the impact of volatility on the market maker's inventory. Furthermore, we find that the stamp duty rate is a critical factor in market making, with a negative impact on both the profit of the market maker and the liquidity of the market. Additionally, our analysis emphasizes the significance of accurately estimating stock drift for managing inventory and adverse selection risks effectively and enhancing profit for the market maker. These findings offer valuable insights for both market makers and policymakers in the Chinese stock market and provide directions for further research in designing effective market making strategies and policies.