LLMs Corrupt Your Documents When You Delegate

2026-04-16

AI Summary

This work addresses the critical yet underexplored issue of silent content corruption introduced by large language models (LLMs) during long-horizon, delegated document editing tasks. To systematically evaluate reliability in such workflows, the authors propose the first assessment framework tailored to delegated AI interactions and introduce DELEGATE-52, a benchmark spanning 52 professional domains. Through comprehensive evaluation of 19 state-of-the-art LLMs, the study reveals that even leading models—such as GPT-5.4 and Claude-4.6 Opus—silently corrupt an average of 25% of document content after extended editing sequences. Moreover, existing agent-based mitigation tools prove ineffective, uncovering a distinct degradation pattern wherein errors remain sparse yet highly destructive, intensifying with both document scale and workflow length.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Safety and Robustness
• Multiagent Systems: Adversarial Agents

Application Category

• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

Abstract

Large Language Models (LLMs) are poised to disrupt knowledge work, with the emergence of delegated work as a new interaction paradigm (e.g., vibe coding). Delegation requires trust - the expectation that the LLM will faithfully execute the task without introducing errors into documents. We introduce DELEGATE-52 to study the readiness of AI systems in delegated workflows. DELEGATE-52 simulates long delegated workflows that require in-depth document editing across 52 professional domains, such as coding, crystallography, and music notation. Our large-scale experiment with 19 LLMs reveals that current models degrade documents during delegation: even frontier models (Gemini 3.1 Pro, Claude 4.6 Opus, GPT 5.4) corrupt an average of 25% of document content by the end of long workflows, with other models failing more severely. Additional experiments reveal that agentic tool use does not improve performance on DELEGATE-52, and that degradation severity is exacerbated by document size, length of interaction, or presence of distractor files. Our analysis shows that current LLMs are unreliable delegates: they introduce sparse but severe errors that silently corrupt documents, compounding over long interaction.

Problem

• Large Language Models
• delegated workflows
• document corruption
• error propagation
• AI reliability

Innovation

• delegated workflows
• document corruption
• LLM reliability
• DELEGATE-52
• error compounding