bilingual-canadian-eval / v1 /code /tasks /ifeval_fr.yaml
Vveerrgg's picture
v1
ed9f38e verified
Raw History Blame Contribute Delete
862 Bytes
# French IFEval (jzhang86/fr_ifeval, Apache-2.0: GPT-4o translation, human-checked); stock IFEval scorer and metrics.
task: ifeval_fr
dataset_path: jzhang86/fr_ifeval
dataset_name: null
output_type: generate_until
test_split: train
num_fewshot: 0
doc_to_text: prompt
doc_to_target: 0
generation_kwargs:
until: []
do_sample: false
temperature: 0.0
max_gen_toks: 1280
process_results: !function utils.ifeval_process_results
metric_list:
- metric: prompt_level_strict_acc
aggregation: mean
higher_is_better: true
- metric: inst_level_strict_acc
aggregation: !function utils.agg_inst_level_acc
higher_is_better: true
- metric: prompt_level_loose_acc
aggregation: mean
higher_is_better: true
- metric: inst_level_loose_acc
aggregation: !function utils.agg_inst_level_acc
higher_is_better: true
metadata:
version: 1.0