Rethinking Multiple-Choice Questions for RLVR: Unlocking Potential via Distractor Design

TL;DR

This paper explores how the design of multiple-choice questions impacts reinforcement learning with verifiable rewards, proposing a new distractor curation method to improve reasoning and training effectiveness.

Contribution

It introduces Iterative Distractor Curation (IDC), a framework for creating high-quality distractors that enhance RLVR training by preventing shortcut reasoning.

Findings

Strong distractors reduce random guessing in RLVR.

Matching option counts between training and testing improves performance.

IDC significantly boosts RLVR training outcomes.

Abstract

Reinforcement Learning with Verifiable Rewards (RLVR) significantly enhances the reasoning capabilities of Large Language Models. When applied to RLVR, Multiple-Choice Questions (MCQs) offer a scalable source of verifiable data but risk inducing reward hacking, where models shortcut reasoning via random guessing or simple elimination. Current approaches often mitigate this by converting MCQs to open-ended formats, thereby discarding the contrastive signal provided by expert-designed distractors. In this work, we systematically investigate the impact of option design on RLVR. Our analysis highlights two primary insights: (1) Mismatches in option counts between training and testing degrade performance. (2) Strong distractors effectively mitigate random guessing, enabling effective RLVR training even with 2-way questions. Motivated by these findings, we propose Iterative Distractor Curation (IDC), a framework that actively constructs high-quality distractors to block elimination shortcuts and promote deep reasoning. Experiments on various benchmarks demonstrate that our method effectively enhances distractor quality and yields significant gains in RLVR training compared to the original data.