Exchange Ranking
Top Exchanges
24h Volume Ranking
Popularity Ranking
Exchange BTC Balance
Proof of Reserves
Decentralized Exchanges
Funding Rate
Funding Heatmap
Liquidation Data
Max Pain
Long/Short Ratio
Whale L/S Ratio
Binance/Okex/Huobi L/S
Bitfinex Margin L/S
ETF Tracker
Solana ETF
XRP ETF
Hong Kong ETF
Bitcoin Treasuries
Crypto Reversal
Ethereum Reserves
HyperLiquid Wallet Analysis
Hyperliquid Whale Watch
Large Transactions
On-chain Movement
Bitcoin ROI
Stablecoin Market Cap
Options Analysis
News
Articles
Economic Calendar
Features
Wallet
Contract Calculator
Security
Collections
Watchlist
Following
ETH Coinbase 推出基于大语言模型的欺诈检测代理，有效降低 Onramp 交易风险
2026年10月6日，Coinbase 发布了一篇题为《Owning Intelligence》系列文章的首篇博文，详细介绍了其 Onramp 产品中新增的基于大语言模型（LLM）的欺诈检测代理。该代理旨在帮助用户将法定货币转换为加密货币的过程中，更精准地识别和拦截欺诈行为。
显著的性能提升与成本效益
在开发过程中，团队将该 LLM 代理与前沿的通用模型进行了基准测试对比。结果显示，新代理在 F1 分数上提升了 9.6%，端到端延迟降低了 55%。与现有基础设施不同，该 LLM 代理并非完全取代 Coinbase 现有的反欺诈系统，而是作为下游组件运行，对早期欺诈管道标记的交易提供“额外的上下文审查”。
为了验证实际效果，Coinbase 进行了一场在线 A/B 测试，将新代理与其现有的机器学习和基于规则的系统进行对比。测试数据显示：
- 欺诈交易数量减少 30%
- 欺诈总金额降低 22%
Coinbase said this result is a significant improvement compared to previous anti-fraud baselines. In addition, the company disclosed that the expected savings achieved by improving anti-fraud efficiency are approximately three times the reasoning cost of running the AI agent. This shows that the deployment is not only more technically efficient, but also economically superior to the required investment of computing resources.
Technical Architecture and Industry Background
In terms of architecture, the system is built on Ray Serve, an open source framework for extending machine learning workloads that reflects Coinbase's infrastructure choices for deploying AI models in production environments.
As crypto exchanges increasingly adopt AI-driven tools to combat fraud involving Onramp and Offramp services, such services have become a common target for criminals trying to introduce illicit funds into or out of the crypto ecosystem through seemingly legal fiat currency conversions. As a result, measurable reductions in fraud rates such as the one reported by Coinbase have become a benchmark that is closely watched across the industry.
This blog post is the first part of Coinbase's "Owning Intelligence" series and hints that the platform plans to detail other in-house AI systems in future articles, covering other parts of its product stack.
