Beyond Reference-Based Evaluation: Reward Models for Meta-Evaluation of Grammatical Error Correction

2026-09-17

AI Summary

本文针对语法纠错评价中参考答案局限性问题，提出基于奖励模型RM-EVAL的方法进行无参考评价，并通过奖励引导文本生成改进纠错系统。

Abstract

Reference-based metrics for Grammatical Error Correction (GEC) such as M$^2$ and ERRANT assume that the reference set enumerates all valid edits, and therefore often penalize corrections that are grammatical and meaning-preserving but phrased differently. We introduce RM-EVAL, a reward model trained on human preference data from SEEDA, as a reference-free meta-evaluator that predicts human-like quality judgments at both full-sequence and partial-sequence levels. Beyond evaluation, we show that the same reward model can be used as a learning signal to improve GEC generation via Reward-Guided Text Generation (RGTG), which keeps a base GEC model frozen and performs online, reward-driven decoding. Across SEEDA, RM-EVAL achieves strong agreement with human rankings, and RGTG yields consistent gains in reward and external validation, demonstrating a unified framework for both assessing and enhancing GEC systems without relying on gold references.

Problem

• Grammatical Error Correction
• Reference-based Metrics
• Human Preference Data

Innovation

• reward model
• reference-free evaluation
• grammatical error correction
• human preference data
• reward-guided text generation