Prompting a Weighting Mechanism into LLM-as-a-Judge in Two-Step: A Case Study

2025-02-19

AI Summary

Large language models (LLMs) employed as natural language generation (NLG) evaluators often suffer from low human alignment due to insufficient awareness of semantic importance—leading them to overemphasize peripheral details while neglecting core meaning. To address this, we propose a zero-shot, two-stage prompting framework: first identifying task-critical semantic dimensions, then explicitly modeling their relative importance via structured, dynamically weighted scoring. Crucially, our method requires no fine-tuning or auxiliary training; instead, it embeds interpretable, importance-aware weighting logic directly into the prompt. This constitutes the first zero-shot evaluation enhancement that explicitly incorporates importance awareness. Evaluated across diverse NLG benchmarks, it improves Human Alignment Rate (HAR) by an average of 6%, significantly mitigating fine-grained evaluation biases. Our approach establishes a lightweight, interpretable, and plug-and-play paradigm for LLM-based NLG assessment.

Technology Category

• Natural Language Processing: (Large) Language Models
• Machine Learning: Large Multimodal Models (LMMs)
• Planning, Routing, and Scheduling: Planning with Language Models

Application Category

• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

While Large Language Models (LLMs) have emerged as promising tools for evaluating Natural Language Generation (NLG) tasks, their effectiveness is limited by their inability to appropriately weigh the importance of different topics, often overemphasizing minor details while undervaluing critical information, leading to misleading assessments. Our work proposes an efficient prompt design mechanism to address this specific limitation and provide a case study. Through strategic prompt engineering that incorporates explicit importance weighting mechanisms, we enhance using LLM-as-a-Judge ability to prioritize relevant information effectively, as demonstrated by an average improvement of 6% in the Human Alignment Rate (HAR) metric.

Problem

• LLM weighting mechanism improvement
• Enhanced NLG task evaluation
• Increased Human Alignment Rate

Innovation

• Two-step prompt design mechanism
• Explicit importance weighting strategy
• Enhanced LLM-as-a-Judge prioritization