windyboy
63add96a09
- Why Guide: remove duplicated 20-case/rubric/test-definition tables, link to authoritative Roadmap/Worksheet instead; 10-min action points to entries - Concept Map / What-Is: cross-link narrative vs quick-reference roles - Worksheet: annotate sections with authoritative sources - 04-Reference 01-04 <-> archive/01: bidirectional resource links - archive/00-Material-List: shrink guidebook listing, now reachable from READMEs - guidebook: compress 06-Yearly-Dives 2025 section (-> 08-2025-Edition §3), add old/new version nav banners to 01-05, reverse links in 08 - README/Start-Here: link Learning Board (was orphaned)
38 lines
1.2 KiB
Markdown
38 lines
1.2 KiB
Markdown
---
|
||
type: reference
|
||
tags:
|
||
- llm-evaluation
|
||
- continuous-evaluation
|
||
- ci
|
||
status: active
|
||
created: 2026-08-21
|
||
---
|
||
|
||
# 03 · Continuous Evaluation
|
||
|
||
主题：把评测从「手工跑一次」升级为「每次变更都会跑」
|
||
|
||
## A 级资源
|
||
|
||
### 1. CircleCI + RAGAS Continuous Eval
|
||
- 价值不在 RAGAS 本身，而在：
|
||
Eval → CI → Threshold → Gate
|
||
- 真正要学习的是：
|
||
Metric → Baseline → Threshold → Action
|
||
例如：
|
||
- security_failure > 0 → fail
|
||
- overall_pass_rate ↓ → review
|
||
- latency regression → review
|
||
- RAGAS 只是 grader/metric 工具
|
||
- 适合阶段：项目已有 regression set 后
|
||
|
||
### 2. DeepLearning.AI + Weights & Biases
|
||
- 真正适合学习的：
|
||
Experiment Tracking / Run / Artifact / Version / Trace / Comparison
|
||
- 建议用途：当你的目录开始出现
|
||
runs/、runs-v2/、runs-final/、runs-final-new/
|
||
说明该学 experiment tracking 了
|
||
- 不建议：一开始就为了"专业"搭 W&B
|
||
|
||
> 🔗 本页资源的完整链接、来源背景与上手建议见 [[04-Reference/archive/01-Curated-External-Resources|archive/01-Curated-External-Resources]]（"云厂商生产环境"与"名校/大牛的工业级课程"两节）。
|