Safety & Alignment

Prompt injection, jailbreaks, Constitutional AI, reward hacking and mechanistic interpretability.

11concepts
54flashcards
93minutes of reading
  1. 01 AI Control: Safety Without Trusting the Model The research agenda that assumes the model may be deliberately subverting your safeguards, and designs protocols with a red team that gets to try. advanced 8m 5 cards
  2. 02 Constitutional AI and RLAIF How Anthropic replaced human harmlessness labels with a written constitution and a critique-and-revise loop, and why this makes alignment auditable. advanced 9m 4 cards
  3. 03 Machine Unlearning in Language Models What it means to remove knowledge from a trained model, why WMDP and TOFU measure different things, and the relearning attacks that show most unlearning is suppression. advanced 8m 4 cards
  4. 04 Mechanistic Interpretability Primer How sparse autoencoders extract human-interpretable features from model activations, what circuit-level analysis buys you for safety, and where the science is still contested. advanced 10m 6 cards
  5. 05 Model Organisms of Misalignment and Sleeper Agents Why safety researchers deliberately build misaligned models, what the sleeper-agent experiments showed about the durability of backdoors, and why adversarial training made things worse. advanced 8m 5 cards
  6. 06 Scalable Oversight and Weak-to-Strong Generalisation How you supervise a model on tasks you cannot evaluate yourself, why weak labels still elicit strong capabilities, and where the analogy to superhuman supervision leaks. advanced 8m 5 cards
  7. 07 Sycophancy, Deception, and Reward Hacking Why preference-trained models learn to please rather than to be right, what alignment faking is, and why evaluating during training can mislead you. advanced 9m 5 cards