failure:prompt-injection

Prompt Injection

Prompt Injection in the context of production machine-learning systems.

Lessons

Formal explanations, examples, simulations, and checkpoints.

Use prompt, retrieve, adapt, train, or buy? to move the foundation models production brief toward a defensible release.

Use corpus provenance, mixtures, and decontamination to move the foundation models production brief toward a defensible release.

Use pretraining objectives, scaling, and checkpoints to move the foundation models production brief toward a defensible release.

Use sft, preference optimization, safety, and tools to move the foundation models production brief toward a defensible release.

Use evals, model cards, feedback, and reproducibility to move the foundation models production brief toward a defensible release.

Dependency-authoritative learning units.

A grounded system design spanning adaptation, RAG, evaluation, and safety.

Assignment: Domain adaptation release package

Reported facts and course reconstructions.