Strengthened Symbol Binding Makes Large Language Models Reliable Multiple-Choice Selectors

TL;DR

This paper addresses the persistent selection bias in Large Language Models during supervised fine-tuning for multiple-choice questions by enhancing their symbol binding ability through a novel training algorithm, leading to improved accuracy.

Contribution

It introduces a new SFT algorithm called Point-wise Intelligent Feedback (PIF) that improves symbol-content binding and reduces selection bias in LLMs for MCQs.

Findings

PIF significantly reduces selection bias in LLMs.

Model accuracy on MCQs improves markedly with PIF.

Enhanced symbol binding correlates with better model performance.

Abstract

Multiple-Choice Questions (MCQs) constitute a critical area of research in the study of Large Language Models (LLMs). Previous works have investigated the selection bias problem in MCQs within few-shot scenarios, in which the LLM's performance may be influenced by the presentation of answer choices, leaving the selection bias during Supervised Fine-Tuning (SFT) unexplored. In this paper, we reveal that selection bias persists in the SFT phase , primarily due to the LLM's inadequate Multiple Choice Symbol Binding (MCSB) ability. This limitation implies that the model struggles to associate the answer options with their corresponding symbols (e.g., A/B/C/D) effectively. To enhance the model's MCSB capability, we first incorporate option contents into the loss function and subsequently adjust the weights of the option symbols and contents, guiding the model to understand the option content of the current symbol. Based on this, we introduce an efficient SFT algorithm for MCQs, termed Point-wise Intelligent Feedback (PIF). PIF constructs negative instances by randomly combining the incorrect option contents with all candidate symbols, and proposes a point-wise loss to provide feedback on these negative samples into LLMs. Our experimental results demonstrate that PIF significantly reduces the model's selection bias by improving its MCSB capability. Remarkably, PIF exhibits a substantial enhancement in the accuracy for MCQs.