Mutable Transcripts: Mitigating Context Pollution through Editable Conversation State

2026-09-25

AI Summary

This study addresses the misalignment between static conversation history and dynamic user intent in existing LLM dialogue systems, where outdated information persistently contaminates the context. To mitigate this, we propose a variable transcription interaction paradigm that reconstructs passively appended dialogue logs into editable state representations, enabling users to revise historical turns via natural language instructions and thereby mechanistically eliminating contextual interference. We developed a prototype system and conducted a controlled user study to evaluate the approach. Results demonstrate that the proposed paradigm significantly outperforms traditional interaction modes in clarity, confidence, and usability. Furthermore, it effectively shortens dialogue length and reduces users’ inclination to restart conversations, achieving precise elimination of obsolete contextual information.

Abstract

Contemporary large language model (LLM) chat systems treat conversation history as an immutable sequence of turns that defines the model's working context. However, user intent in real interactions is not static: it evolves through correction, refinement, and shifting constraints. This mismatch between dynamic intent and static transcripts can result in context pollution, where outdated or irrelevant information persists and continues to influence subsequent responses. We introduce mutable transcripts, a new interaction paradigm that enables users to revise prior turns through natural language edit requests, allowing the conversation history itself to be updated rather than appended. This reframes the transcript from a passive record into an editable representation of conversational state. We present a working prototype that integrates transcript-level revision into a standard chat interface and evaluate its feasibility through a controlled user study (n=17) and an illustrative transcript analysis of representative interaction scenarios. Participants significantly preferred mutable transcripts over standard chat across measures of clarity, confidence, and ease of use, with reduced intent to restart conversations. Transcript analysis of representative user study conversations shows that mutable transcripts can reduce conversation length and eliminate obsolete retained context. These findings provide initial evidence that user-driven revision of conversational history can improve interaction quality and help maintain a more current representation of user intent. The source code and prototype can be accessed at https://github.com/QxLabIreland/ReChat

Problem

• context pollution
• mutable transcripts
• conversation state
• large language models
• dynamic user intent

Innovation

• Mutable Transcripts
• Context Pollution
• Editable Conversation State
• Large Language Models
• Natural Language Editing