About reward function

[LaurenceSu1037]

VerIF/verl/utils/reward_score/__init__.py

Line 17 in 5c86f57

def _default_compute_score(prompt_str, data_source, solution_str, ground_truth, extra_info=None):

Thanks for your work! I would like to know where can I find the "prompt_str"? It is different from Verl(verl-dev-0.5.0) I used? So which part is the prompt_str I can attach in if_prompt.py?

[h-peng17]

"prompt_str" is encoded from

VerIF/examples/data_preprocess/if_prompts.py

Line 65 in 5c86f57

"prompt": [

.
Verl will apply chat template and encode "prompt" to "prompt_str".