Standards used in this guide

The recipes use published formats and tools wherever one exists. Where the program's workstreams provide measurements or tools, the guide uses them.

Model formats, cards, and licences

Standard	Use	In this guide
Hugging Face model cards and Mitchell et al. (2019)	The card format and sections a model needs before the organization considers it	Chapter 2
Apache 2.0, MIT, and Responsible AI Licenses	The licence families the classification distinguishes	Chapter 2
safetensors and GGUF	Weight formats that load without executing code	Chapters 3 and 6
Sigstore model transparency	Signing and verifying model files	Chapter 6

Serving and adaptation

Tool	Use	In this guide
vLLM, llama.cpp, Ollama	Serving stacks from workstation to shared service	Chapter 3
Sentence Transformers and the PI-FT toolkit	Embedding models and their fine-tuning on metadata	Chapter 4
Hugging Face PEFT	Parameter-efficient fine-tuning of small generators	Chapter 4

Evaluation and governance

Resource	Use	In this guide
Evaluation suites cookbook	Suites, required scores, gates, report cards	Chapters 1, 4, 5, and 6
lm-evaluation-harness, Open LLM Leaderboard, MTEB	Public benchmarks for shortlisting	Chapters 2 and 5
Efficient and Inclusive AI Applications	The program's measurements of model size against task accuracy and its deployment guidance	Chapters 1, 3, 5, and 7
OWASP Top 10 for LLM applications	Supply chain and deployment risks	Chapter 6
AI component register	Where choices, versions, and shared assets are recorded	All chapters

Corrections

Standards change versions and URLs. If a reference here is wrong or out of date, open an issue with the row and the corrected source.