LLM as a Broken Telephone: Iterative Generation Distorts Information

Amr Mohamed, Mingmeng Geng, Michalis Vazirgiannis, Guokan Shang
2/27/2025
cs.CLcs.AI

No official code found. Best implementation: amr-mohamedd/LLM-as-a-Broken-Telephone (9 stars, last commit 2025-03-11).

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.

View on arXivView PDF

Code Implementations(1)

amr-mohamedd/LLM-as-a-Broken-TelephoneCommunity100%

[ACL 2025] Code implementation for the "LLM as a Broken Telephone: Iterative Generation Distorts Information" paper.

90PythonFeb 27, 20251 years ago
agentic-aigenerationinferencellmllm-inference

Similar Papers

Understanding LLM Failures: A Multi-Tape Turing Machine Analysis of Systematic Errors in Language Model Reasoning

84% match2026

Bias Amplification in Language Model Evolution: An Iterated Learning Perspective

83% match1 repo52024

Security Degradation in Iterative AI Code Generation -- A Systematic Analysis of the Paradox

83% match2025

Beyond speculation: Measuring the growing presence of LLM-generated texts in multilingual disinformation

83% match2025

When LLMs Play the Telephone Game: Cultural Attractors as Conceptual Tools to Evaluate LLMs in Multi-turn Settings

83% match1 repo122024

When LLMs Play the Telephone Game: Cultural Attractors as Conceptual Tools to Evaluate LLMs in Multi-turn Settings

83% match1 repo122024

Cite this paper

@article{mohamed2025broken,
  title  = {LLM as a Broken Telephone: Iterative Generation Distorts Information},
  author = {Amr Mohamed and Mingmeng Geng and Michalis Vazirgiannis and Guokan Shang},
  year   = {2025},
  eprint = {2502.20258},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2502.20258v2}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2502.20258.svg)](https://codeofpaper.com/paper/2502.20258)

Discussion