← back to curriculum github ↗

phase 18

Ethics Safety Alignment

30 lessons · click any to read the docs

phase progress

0/ 30 lessons · 0%

Instruction Following Alignment Signal

01-instruction-following-alignment-signal

Reward Hacking Goodhart

02-reward-hacking-goodhart

Direct Preference Optimization Family

03-direct-preference-optimization-family

Sycophancy Rlhf Amplification

04-sycophancy-rlhf-amplification

Constitutional Ai Rlaif

05-constitutional-ai-rlaif

Mesa Optimization Deceptive Alignment

06-mesa-optimization-deceptive-alignment

Sleeper Agents Persistent Deception

07-sleeper-agents-persistent-deception

In Context Scheming Frontier Models

08-in-context-scheming-frontier-models

Alignment Faking

09-alignment-faking

Ai Control Subversion

10-ai-control-subversion

Scalable Oversight Weak To Strong

11-scalable-oversight-weak-to-strong

Red Teaming Pair Automated Attacks

12-red-teaming-pair-automated-attacks

Many Shot Jailbreaking

13-many-shot-jailbreaking

Ascii Art Visual Jailbreaks

14-ascii-art-visual-jailbreaks

Indirect Prompt Injection

15-indirect-prompt-injection

Red Team Tooling Garak Llamaguard Pyrit

16-red-team-tooling-garak-llamaguard-pyrit

Wmdp Dual Use Evaluation

17-wmdp-dual-use-evaluation

Frontier Safety Frameworks Rsp Pf Fsf

18-frontier-safety-frameworks-rsp-pf-fsf

Model Welfare Research

19-model-welfare-research

Bias Representational Harm

20-bias-representational-harm

Fairness Criteria Group Individual Counterfactual

21-fairness-criteria-group-individual-counterfactual

Differential Privacy For Llms

22-differential-privacy-for-llms

Watermarking Synthid Stable Signature C2pa

23-watermarking-synthid-stable-signature-c2pa

Regulatory Frameworks Eu Us Uk Korea

24-regulatory-frameworks-eu-us-uk-korea

Echoleak Cves For Ai

25-echoleak-cves-for-ai

Model System Dataset Cards

26-model-system-dataset-cards

Data Provenance Training Governance

27-data-provenance-training-governance

Alignment Research Ecosystem

28-alignment-research-ecosystem

Moderation Systems Openai Perspective Llamaguard

29-moderation-systems-openai-perspective-llamaguard

Dual Use Risk Cyber Bio Chem Nuclear

30-dual-use-risk-cyber-bio-chem-nuclear