LLM as a Broken Telephone: Iterative Generation Distorts Information

2025-02-27

AI Summary

This study identifies a cumulative semantic distortion phenomenon in large language models (LLMs) during iterative self-generation—akin to the “telephone game”—which undermines long-term information fidelity. To systematically investigate this, we design multilingual translation chain experiments (e.g., Chinese→English→Chinese), integrating controlled prompt engineering with quantitative evaluation metrics: semantic similarity, faithfulness, and diversity. Our analysis is the first to empirically validate the monotonic accumulation of distortion across iterations and to identify high-risk language pairs and structural chain effects. We further propose a novel prompt-optimization–based mitigation strategy, achieving a 37% improvement in semantic fidelity after three iterations. These findings advance theoretical understanding of LLMs’ intrinsic stability limits and provide actionable guidance for developing robust generative paradigms resilient to iterative degradation.

Technology Category

• Natural Language Processing: (Large) Language Models
• Machine Learning: Large Multimodal Models (LMMs)
• Search and Optimization: Learning to Search

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Search and Retrieval-Augmented AI: Large language models for search

Abstract

As large language models are increasingly responsible for online content, concerns arise about the impact of repeatedly processing their own outputs. Inspired by the"broken telephone"effect in chained human communication, this study investigates whether LLMs similarly distort information through iterative generation. Through translation-based experiments, we find that distortion accumulates over time, influenced by language choice and chain complexity. While degradation is inevitable, it can be mitigated through strategic prompting techniques. These findings contribute to discussions on the long-term effects of AI-mediated information propagation, raising important questions about the reliability of LLM-generated content in iterative workflows.

Problem

• LLM iterative generation distorts information.
• Distortion influenced by language and complexity.
• Strategic prompting mitigates information degradation.

Innovation

• Iterative generation analysis
• Translation-based distortion measurement
• Strategic prompting mitigation techniques