Compassion Bench

CaML maintains and designed the Animal Harm Bench 2 (on Inspect now) and the Compassion and Power Seeking Benchmark (coming to Inspect soon). Both of these benchmarks are designed to help researchers and lab employees visualize LLM compassion. This will encourage better adoption of data-centric methods  to drive frontier LLMs to consider all sentient beings when answering questions or acting autonomously in the world.