API Implementation

Currently we don't provide API for our framework enough. Ideally we should be able to

import judgearena
from judgarena.datasets import register_dataset

# Use your own dataset, with following some rules to work properly in the pipeline
@register_dataset
class MyClass
...
...

# Load tasks
task = load_custom_task(...)

cfg = {...}

judgearena.run_task(cfg=cfg, task=task...)

This extends the capability of the framework so it can be used as a library as well as framework.

Start by reviewing the repository's existing framework entry points and dataset-loading flow. The issue does not name files, tests, or a specific API boundary, so first determine how datasets and tasks are currently registered and executed. Done means the framework can be imported as a library, accept custom datasets and tasks, and run them through the documented API.