DATA CUT 2026-07-27 116 活跃模型 680 A 类案例

CASE EVIDENCE / A RECORD

untanao 使用 Kimi K2 处理软件工程任务执行

untanao 公开的代码代理与软件工程案例,来源为 公开代码库,复核于 2026-06-27T05:52:39Z。

原始记录:Token DD Agent uses Kimi K2 as the default reasoning model for crypto token due diligence

A

Chinese Brief

中文案例导读

untanao 公开的代码代理与软件工程案例,来源为 公开代码库,复核于 2026-06-27T05:52:39Z。 Model Atlas 将它标记为 A 类证据,因为它同时具备具体使用者、具体任务、公开原始证据和可访问产物。 Model Atlas 不把 benchmark、教程、发布说明或集合页包装成真实案例。

厂商Kimi / Moonshot AI
模型Kimi K2
任务类型代码代理与软件工程
审核状态auto_approved

任务

真实任务背景

这是一个围绕软件工程任务执行的真实任务,公开材料可以回溯到具体使用者和具体产物。 原始资料写作:Open-source web agent where a user pastes an EVM token contract address and the system gathers Etherscan metadata, DexScreener market data, GoPlus security signals, honeypot simulation results, and web evidence to asses…

代码代理与软件工程公开代码库A 类可核验real_case
公开产物

公开材料提供公开代码、README 或项目配置,可用于核验任务结果、项目形态和模型绑定关系。 原始资料写作:The application streams tool-call progress to the UI and produces a structured Markdown due-diligence report with a 0-100 risk score and verdict for the submitted token.

模型作用

Kimi K2 在该案例中承担软件工程任务执行相关的生成、分析、编排或实现角色。 原始资料写作:The README states Kimi K2 works out of the box and is the default reasoning model via Fireworks (`accounts/fireworks/models/kimi-k2p6`). Kimi K2 drives the agent loop, selects among available tools, interprets retrieved…

风险边界

当前判断基于公开材料;若产物下线、仓库变更或模型参与比例仅来自作者自述,需要在引用前重新复核。 原始资料写作:Evidence is from the public GitHub README and code artifact; the project is model-agnostic and supports other LLMs, but its documented default model is Kimi K2 via Fireworks.