How Do Language Models Process Ethical Instructions? Deliberation, Consistency, and Other-Recognition Across Four Models

Create an automated multi-agent framework to test how different alignment techniques and instructions influence ethical behavior under pressure. This is a must-have for safety researchers developing 'hardened' LLMs.

Suggested repo: align-sim

"Stress-test your model's ethical alignment using multi-agent simulations."

Estimated effort: 45h