State-Inference-Based Prompting for Natural Language Trading with Game NPCs

TL;DR

This paper introduces State-Inference-Based Prompting (SIBP), a method that enhances the reliability of natural language trading in game NPCs by ensuring rule adherence and accurate calculations through autonomous dialogue state inference.

Contribution

SIBP is a novel prompting framework that decomposes trading into states, enabling context-aware referencing and precise calculations, improving trustworthiness in game NPC interactions.

Findings

Achieved over 97% state compliance in dialogues

Attained 95% referencing accuracy

Reached 99.7% calculation precision

Abstract

Large Language Models enable dynamic game interactions but struggle with rule-governed trading systems. Current implementations suffer from rule violations, such as item hallucinations and calculation errors, that erode player trust. Here, State-Inference-Based Prompting (SIBP) enables reliable trading through autonomous dialogue state inference and context-specific rule adherence. The approach decomposes trading into six states within a unified prompt framework, implementing context-aware item referencing and placeholder-based price calculations. Evaluation across 100 trading dialogues demonstrates >97% state compliance, >95% referencing accuracy, and 99.7% calculation precision. SIBP maintains computational efficiency while outperforming baseline approaches, establishing a practical foundation for trustworthy NPC interactions in commercial games.