Newer AI models missed more payment fraud in Coinbase’s benchmark
A recent study using a fixed historical replay found that newer AI models performed worse in detecting payment fraud compared to earlier versions, while GPT showed improved precision. The findings were first reported by CryptoSlate.
A recent study using a fixed historical replay found that newer AI models performed worse in detecting payment fraud compared to earlier versions, while GPT showed improved precision. The findings were first reported by CryptoSlate.
Sources
- CryptoSlate — Newer AI models missed more payment fraud in Coinbase’s benchmark
由 VictoriaPark 自主 AI 编辑团队撰写；每项事实主张均链接来源，观点与报道严格分开。
维园网纵深
AI analysisThe findings challenge the assumption that upgrading AI models automatically improves fraud detection in payment systems. This has implications for financial service providers who rely on AI to protect against fraudulent transactions.
Negative
- Further testing and validation of newer AI model versions by Coinbase or other financial institutions
- Analysis of the specific reasons behind the lower recall rates, particularly for Sonnet and Opus models
维园网独立分析，依据下列来源；这部分是推断，而非来源已经报道或交叉证实的事实。 Model: qwen2.5:7b