Benchmarking Temporal Web3 Intelligence: Lessons from the FinSurvival 2025 Challenge

2026-02-26

AI Summary

This work addresses the limitation of existing temporal Web3 intelligence benchmarks, which fail to model real-world time dynamics such as censoring and non-stationarity, thereby hindering methodological advancement and cross-domain transfer. Leveraging 21.8 million on-chain transactions from the Aave v3 protocol, we introduce the FinSurvival 2025 Challenge, comprising 16 survival prediction tasks designed to capture the temporal evolution of user behavior. We innovatively treat Web3 as a high-fidelity sandbox and propose a domain-informed temporal feature engineering approach, alongside a reproducible benchmark framework. Experimental results demonstrate that domain-driven temporal features significantly outperform generic models, establishing a new design paradigm and empirical foundation for next-generation temporal intelligence benchmarks.

Technology Category

• Data Mining & Knowledge Management: Mining of Spatial, Temporal or Spatio-Temporal Data
• Application Domains: Humanities & Computational Social Science
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• Economics, Online Markets and Human Computation: Sustainability of Web economics
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Web Mining and Content Analysis: Web data provenance, reliability, and authenticity

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 transitions.We detail the benchmark design and the winning solutions, highlighting how domain-aware temporal feature construction significantly outperformed generic modeling approaches. Furthermore, we distill lessons for next-generation temporal benchmarks, arguing that Web3 systems provide a high-fidelity sandbox for studying temporal challenges, such as churn, risk, and evolution that are fundamental to the wider Web.

Problem

• Temporal Web3
• benchmarking
• non-stationarity
• censoring
• longitudinal data

Innovation

• Temporal Web3
• survival prediction
• benchmark design
• domain-aware features
• non-stationarity