Skip to content

mmlu_pro score issue (generate_prompt) #108

@codemobi

Description

@codemobi

mmlu_pro score is very low for Qwen2.5-32B-Instruct (8.01).

When I changed generate_prompt() in MMLUProTaskHandler class like generate_prompt() MMLUTaskHandler class, the score was increased into 57.80.

skythought/evals/tasks/mmlu/mmlu_handler.py

class MMLUProTaskHandler(MMLUTaskHandler):
def generate_prompt(self, prompt):
multiple_choice_string = self.get_multiple_choice_answers(prompt) # ADDED
prompt = prompt["question"] + "\n" + multiple_choice_string # ADDED
return self.task_config.templating_parameters["template"].format(prompt=prompt)

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type
    No fields configured for issues without a type.

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions