restructure 04-Reference: resource map for evaluation engineering (archive old drafts)
This commit is contained in:
@@ -0,0 +1,55 @@
|
||||
---
|
||||
type: reference
|
||||
tags:
|
||||
- llm-evaluation
|
||||
- evaluation-infrastructure
|
||||
status: active
|
||||
created: 2026-08-21
|
||||
---
|
||||
|
||||
# 01 · Evaluation Infrastructure
|
||||
|
||||
主题:Evaluation Harness、Sandbox、Trace、Scaling
|
||||
|
||||
## S 级资源(必须认真研究)
|
||||
|
||||
### 1. UK AISI Engineering Playbook + Inspect AI
|
||||
- 仓库:https://github.com/UKGovernmentBEIS/inspect_ai
|
||||
- 背景:英国人工智能安全研究所(UK AISI)官方开源
|
||||
- 为什么看:国家级安全评测机构测试前沿模型时使用的完整底座
|
||||
- 核心思想:把评测基础设施拆成五层
|
||||
- Evaluate
|
||||
- Isolate
|
||||
- Connect
|
||||
- Run
|
||||
- Scale
|
||||
- 重点抽象(映射到自己的体系):
|
||||
- Task → Case
|
||||
- Dataset → Dataset
|
||||
- Solver → Model / Agent Adapter
|
||||
- Tool / Sandbox → 隔离执行环境
|
||||
- Scorer → Grader
|
||||
- Log → Trace / Outcome
|
||||
- 适合阶段:完成第一个小项目以后
|
||||
|
||||
### 2. AWS Generative AI Evaluations Workshop
|
||||
- 为什么看:目前垂直场景最全、最硬核的可运行实战代码
|
||||
- 覆盖场景:
|
||||
- Multimodal RAG
|
||||
- Tool Calling(5 种渐进式评测方法)
|
||||
- Automated Reasoning(利用 SMT 求解器检查合规)
|
||||
- Multi-Agent Shared Context
|
||||
- Red Teaming
|
||||
- 学习方式(重要):
|
||||
不要只照着 Notebook 跑。每个模块都问:
|
||||
- Task 是什么?
|
||||
- Case 怎么构造?
|
||||
- Rubric 是什么?
|
||||
- Grader 是什么?
|
||||
- Failure 如何定义?
|
||||
- 如何做成 Regression?
|
||||
- 适合阶段:最适合作为第一个实操资源
|
||||
|
||||
## 次级参考
|
||||
- Hugging Face evaluation-guidebook(已在本目录下:[[04-Reference/evaluation-guidebook/00-Overview|evaluation-guidebook]])
|
||||
- DeepEval(应用级单元测试框架,上手快但抽象层较浅)
|
||||
Reference in New Issue
Block a user