M-MAD: Multidimensional Multi-Agent Debate for Advanced Machine Translation Evaluation

TL;DR

M-MAD introduces a multi-agent debate framework utilizing LLMs for detailed, reliable machine translation evaluation, outperforming existing LLM-based methods and rivaling state-of-the-art automatic metrics.

Contribution

It presents a novel multi-agent debate approach that decouples evaluation criteria and synthesizes results, advancing LLM-based MT evaluation.

Findings

Outperforms existing LLM-as-a-judge methods

Competes with state-of-the-art automatic metrics

Demonstrates robustness with suboptimal models

Abstract

Recent advancements in large language models (LLMs) have given rise to the LLM-as-a-judge paradigm, showcasing their potential to deliver human-like judgments. However, in the field of machine translation (MT) evaluation, current LLM-as-a-judge methods fall short of learned automatic metrics. In this paper, we propose Multidimensional Multi-Agent Debate (M-MAD), a systematic LLM-based multi-agent framework for advanced LLM-as-a-judge MT evaluation. Our findings demonstrate that M-MAD achieves significant advancements by (1) decoupling heuristic MQM criteria into distinct evaluation dimensions for fine-grained assessments; (2) employing multi-agent debates to harness the collaborative reasoning capabilities of LLMs; (3) synthesizing dimension-specific results into a final evaluation judgment to ensure robust and reliable outcomes. Comprehensive experiments show that M-MAD not only outperforms all existing LLM-as-a-judge methods but also competes with state-of-the-art reference-based automatic metrics, even when powered by a suboptimal model like GPT-4o mini. Detailed ablations and analysis highlight the superiority of our framework design, offering a fresh perspective for LLM-as-a-judge paradigm. Our code and data are publicly available at https://github.com/SU-JIAYUAN/M-MAD.