ChoiceBench is a lightweight framework for MCQ evaluation-method research on LLMs, with built-in support for answer-order bias analysis and mitigation methods.
python benchmarking research asyncio multiple-choice positional-bias llm prompting mmlu arc-challenge
-
Updated
Aug 7, 2026 - Python