Shiwei Lyu

Most start-ups fail, and early-stage ventures face even lower survival rates. Identifying high-potential start-ups remains a critical challenge for venture capital (VC) investors and policymakers. While predictive models exist, the evolving relationships between VC investors, start-ups, and management teams in dynamic networks are underexplored. We propose a method to predict whether a start-up will succeed within 5 years of its first funding round. Using a 40-year global VC dataset, we model the VC ecosystem as a dynamic bipartite network linking start-ups to individuals (investors/managers). Our approach incrementally updates graph embeddings through unsupervised self-attention to incorporate new nodes, edges, and their neighbors. Node embeddings are further fine-tuned via link prediction and classification tasks, while temporal dependencies are captured to form sequential representations. The model identifies early-stage start-ups with twice the success likelihood of those chosen by professional investors. Key factors including networking and education align with VC literature. Additionally, we provide model complexity analysis and open source our implementation to support practical applications and future research.

Large language models (LLMs) are gaining increasing interests to improve clinical efficiency, owing to their unprecedented performance in modelling natural language. Ensuring the reliable clinical applications, the evaluation of LLMs indeed becomes critical for better mitigating the potential risks, e.g., hallucinations. However, current evaluation methods heavily rely on labor-intensive human participation to achieve human-preferred judgements. To overcome this challenge, we propose an automatic evaluation paradigm tailored to assess the LLMs' capabilities in delivering clinical services, e.g., disease diagnosis and treatment. The evaluation paradigm contains three basic elements: metric, data, and algorithm. Specifically, inspired by professional clinical practice pathways, we formulate a LLM-specific clinical pathway (LCP) to define the clinical capabilities that a doctor agent should possess. Then, Standardized Patients (SPs) from the medical education are introduced as the guideline for collecting medical data for evaluation, which can well ensure the completeness of the evaluation procedure. Leveraging these steps, we develop a multi-agent framework to simulate the interactive environment between SPs and a doctor agent, which is equipped with a Retrieval-Augmented Evaluation (RAE) to determine whether the behaviors of a doctor agent are in accordance with LCP. The above paradigm can be extended to any similar clinical scenarios to automatically evaluate the LLMs' medical capabilities. Applying such paradigm, we construct an evaluation benchmark in the field of urology, including a LCP, a SPs dataset, and an automated RAE. Extensive experiments are conducted to demonstrate the effectiveness of the proposed approach, providing more insights for LLMs' safe and reliable deployments in clinical practice.

When a customer sees a movie recommendation, she may buy the ticket right away, which is the immediate feedback that helps improve the recommender system. Alternatively, she may choose to come back later and this long-term feedback is also modeled to promote user retention. However, the long-term feedback comes with non-trivial challenges in understanding user retention: the complicated correlation between current demands and follow-up demands, coupled with the periodicity of services. For instance, before the movie, the customer buys popcorn through the App, which temporally correlates with the initial movie recommendation. Days later, she checks the App for new movies, as a weekly routine. To address this complexity in a more fine-grained revisit modeling, we propose Time Aware Service Sequential Recommendation (TASSR) for user retention, which is equipped with a multi-task design and an In-category TimeSeqBlock module. Large-scale online and offline experiments demonstrate its significant advantages over competitive baselines.

Unexpected incidents can be destructive or even disastrous, affecting the financial markets. Incidents such as the 9/11 attacks (2001), the Fukushima nuclear disaster (2011), and the COVID-19 outbreaks (2019, 2020) severely shocked both local and global markets. For investors, it is crucial to quantify the key facts that affect the incidents' impacts, and to estimate the reactions of the markets accurately and efficiently for global event-driven investment strategies. Though Web data and other alternative data allow such a possibility, it is still very challenging to mine noisy and often biased heterogeneous data sources, and construct a unified framework for modeling global markets across across time and regions. As a first attempt, we build a framework that extracts incident facts globally based on a deep neural network, feeds them into models built on a global event database complemented with novel socioeconomic datasets (e.g. nightlight data from satellites), and predicts stock market directions in a simulated real-world setting with interpretable results that outperform various baselines. Specifically, we study terrorist attacks in three countries for over 20 years on average, as a first effort to systematically quantify the impact on stock markets at a large scale using novel indicators.