![[2602.23159] Benchmarking Temporal Web3 Intelligence: Lessons from the FinSurvival 2025 Challenge](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.23159] Benchmarking Temporal Web3 Intelligence: Lessons from the FinSurvival 2025 Challenge
Summary
The paper presents the FinSurvival 2025 Challenge, focusing on benchmarking temporal Web3 intelligence using 21.8 million transaction records from the Aave v3 protocol to enhance understanding of user behavior over time.
Why It Matters
As the field of Web3 evolves, establishing reproducible benchmarks is crucial for advancing methodologies in temporal analytics. This study highlights the importance of domain-specific features in improving predictive models, which can significantly impact the development of decentralized applications and user engagement strategies.
Key Takeaways
- The FinSurvival 2025 Challenge provides a framework for benchmarking temporal Web3 intelligence.
- Domain-aware temporal feature construction outperforms generic modeling approaches in survival prediction tasks.
- The study emphasizes the need for shared benchmarks to facilitate methodological progress in Web3 analytics.
- 21.8 million transaction records were utilized to model user behavior effectively.
- Web3 systems serve as high-fidelity environments for studying temporal dynamics like churn and risk.
Computer Science > Machine Learning arXiv:2602.23159 (cs) [Submitted on 26 Feb 2026] Title:Benchmarking Temporal Web3 Intelligence: Lessons from the FinSurvival 2025 Challenge Authors:Oshani Seneviratne, Fernando Spadea, Adrien Pavao, Aaron Micah Green, Kristin P. Bennett View a PDF of the paper titled Benchmarking Temporal Web3 Intelligence: Lessons from the FinSurvival 2025 Challenge, by Oshani Seneviratne and Fernando Spadea and Adrien Pavao and Aaron Micah Green and Kristin P. Bennett View PDF HTML (experimental) Abstract:Temporal Web analytics increasingly relies on large-scale, longitudinal data to understand how users, content, and systems evolve over time. A rapidly growing frontier is the \emph{Temporal Web3}: decentralized platforms whose behavior is recorded as immutable, time-stamped event streams. Despite the richness of this data, the field lacks shared, reproducible benchmarks that capture real-world temporal dynamics, specifically censoring and non-stationarity, across extended horizons. This absence slows methodological progress and limits the transfer of techniques between Web3 and broader Web domains. In this paper, we present the \textit{FinSurvival Challenge 2025} as a case study in benchmarking \emph{temporal Web3 intelligence}. Using 21.8 million transaction records from the Aave v3 protocol, the challenge operationalized 16 survival prediction tasks to model user behavior this http URL detail the benchmark design and the winning solutions, highlight...
