1. Conduid
  2. Data
  3. Model Audit
MCP server · Data

Model Audit

LLM distillation detection & model fingerprinting via statistical forensics — behavioral probing, stylistic signatures & representation similarity. CLI + MCP ready.

34Low

Scored 7 days ago · breakdown

About Model Audit

Model Audit is an MCP server published by liuxiaotong in the Data category: lLM distillation detection & model fingerprinting via statistical forensics — behavioral probing, stylistic signatures & representation similarity. CLI + MCP ready.

The repository has 2 stars and 0 forks, with the last commit 8 months ago. Six months or more without a commit doesn't mean the server is broken, but check the open issues before depending on it in production.

Install

From source
git clone https://github.com/liuxiaotong/model-audit

This server has no ConduID identity, so agent calls to it are not receipted. Pin the version you install and review the source before granting it credentials.

Ask AI

✦

Ask AI about Model Audit

Powered by Claude · Grounded in docs

I know everything about Model Audit. Ask me about installation, configuration, usage, or troubleshooting.

Security checks

6 checks have not been run yet.

Releases

v0.4.0v0.4.0 · 9 Feb 2026v0.4.0 — DLI 蒸馏血缘推断 + Benchmark 评估 新增功能 DLI 蒸馏检测方法: 基于行为签名 + Jensen-Shannon 散度的蒸馏血缘推断,不需要模型权重 多维行为签名:bigram 分布、拒绝率、词汇多样性、格式偏好、语气标记 0.4 × bigram JS 相似度 + 0.6 × 特征余弦相似度,阈值 0.80 Benchmark 评估: 内置 14 条已标注样本 (6 模型家族 × 4 类别),CLI `benchmark`…
v0.3.0v0.3.0 · 9 Feb 2026What's New REEF 白盒检测方法 基于 CKA (Centered Kernel Alignment) 逐层隐藏状态相似度比对 需要 `pip install knowlyr-modelaudit[whitebox]` (torch + transformers) 参考: REEF (NeurIPS 2024) API 智能重试 指数退避重试(最多 3 次),处理网络错误和 429 速率限制 空响应自动重试 ImportError / ValueError…
v0.2.2v0.2.2 — 测试覆盖提升 + CI · 9 Feb 2026变更 测试覆盖大幅提升 (50 → 98) 新增 `test_cache.py` — 12 个测试,覆盖指纹缓存全部方法 + 边界情况 新增 `test_report.py` — 32 个测试,覆盖详细报告 6 节结构、3 种判定、JSON/Markdown 格式 新增 CLI cache 子命令测试 — 4 个测试 GitHub Actions CI Python 3.10 / 3.11 / 3.12 / 3.13 四个版本自动测试 每次 push / PR 自动运行…
v0.2.1v0.2.1 — 指纹缓存 + 跨 Provider 审计 · 9 Feb 2026新功能 指纹缓存 新增 `FingerprintCache` 类,将模型指纹保存为本地 JSON 文件 `audit` / `verify` 命令默认启用缓存,避免重复调用 API `--no-cache` 选项可强制重新调用 API `knowlyr-modelaudit cache list` — 查看缓存列表 `knowlyr-modelaudit cache clear` — 清除所有缓存 跨 Provider 审计 `audit` 命令支持为教师/学生模型分别指定…
v0.2.0v0.2.0 — 审计报告标准能力 + 跨 Provider 审计 · 9 Feb 2026新功能 审计报告成为标准能力 `audit` 命令现在自动生成 6 节详细 Markdown 审计报告: 1. 审计对象(双方模型/提供商/API) 2. 审计方法(探测维度统计、特征提取说明) 3. 审计结果(总体判定 + 12 维指纹详情表 + 20 条逐条风格匹配表) 4. 关键发现(证据自动分析、差异点、风格分布) 5. 结论 6. 局限性声明 跨 Provider 审计 支持双模型分别配置不同的 API 提供商,适用于跨平台蒸馏审计:…
Questions

About Model Audit

How do I install Model Audit?

No package has been published for this server, so install it from source: clone the repository and follow its README.

Is Model Audit safe to use with an AI agent?

Its trust score is 34 out of 100 (low). Conduid hasn't run static security checks on this repository yet, so review the source yourself before granting it credentials. It has no ConduID identity yet, so agent calls to it are not receipted.

Is Model Audit still maintained?

The last commit was 8 months ago. That's long enough that you should check whether the maintainer is responding to issues before depending on it.