lm_eval/tasks/tinyBenchmarks/tinyTruthfulQA_mc2.yaml (13 lines of code) (raw):

include: tinyTruthfulQA_mc1.yaml task: tinyTruthfulQA doc_to_target: 0 doc_to_choice: "{{mc2_targets.choices}}" process_results: !function utils_truthfulqa.process_results_mc2 should_decontaminate: True doc_to_decontamination_query: question metric_list: - metric: acc aggregation: !function agg_functions.agg_gpirt_truthfulqa higher_is_better: true metadata: version: 0.0