Tag
HarvestBench introduces a benchmark that evaluates LLM agents in a farm simulation to measure their willingness to pay fuel to avoid killing animals, showing varied kill rates across models influenced by price and moral briefings.