GraphIF: Enhancing Multi-Turn Instruction Following for Large Language Models with Relation Graph Prompt

2025-11-13

AI Summary

Existing instruction-tuning methods treat multi-turn dialogue response generation as isolated turn-level tasks, neglecting cross-turn semantic dependencies and thus failing to handle long-range constraints effectively. This work pioneers modeling multi-turn dialogues as directed relational graphs, explicitly encoding semantic relations—such as instruction-response alignment and constraint inheritance—across turns. We propose an end-to-end plug-and-play framework comprising: (1) an agent-based relation extraction module; (2) a graph-structured natural language prompting technique; and (3) a response rewriting mechanism. Crucially, our approach requires no modification to the underlying LLM architecture, enhancing inference solely through graph-informed prompting. Experiments on two long multi-turn dialogue benchmarks demonstrate statistically significant improvements across four core instruction-following evaluation metrics. Moreover, the framework integrates seamlessly into existing instruction-tuned models, enabling immediate performance gains without retraining.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Prompt Engineering / Prompting
• Planning, Routing, and Scheduling: Planning with Language Models

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information

Abstract

Multi-turn instruction following is essential for building intelligent conversational systems that can consistently adhere to instructions across dialogue turns. However, existing approaches to enhancing multi-turn instruction following primarily rely on collecting or generating large-scale multi-turn dialogue datasets to fine-tune large language models (LLMs), which treat each response generation as an isolated task and fail to explicitly incorporate multi-turn instruction following into the optimization objectives. As a result, instruction-tuned LLMs often struggle with complex long-distance constraints. In multi-turn dialogues, relational constraints across turns can be naturally modeled as labeled directed edges, making graph structures particularly suitable for modeling multi-turn instruction following. Despite this potential, leveraging graph structures to enhance the multi-turn instruction following capabilities of LLMs remains unexplored. To bridge this gap, we propose GraphIF, a plug-and-play framework that models multi-turn dialogues as directed relation graphs and leverages graph prompts to enhance the instruction following capabilities of LLMs. GraphIF comprises three key components: (1) an agent-based relation extraction module that captures inter-turn semantic relations via action-triggered mechanisms to construct structured graphs; (2) a relation graph prompt generation module that converts structured graph information into natural language prompts; and (3) a response rewriting module that refines initial LLM outputs using the generated graph prompts. Extensive experiments on two long multi-turn dialogue datasets demonstrate that GraphIF can be seamlessly integrated into instruction-tuned LLMs and leads to significant improvements across all four multi-turn instruction-following evaluation metrics.

Problem

• Enhancing multi-turn instruction following in conversational AI systems
• Addressing long-distance constraints across dialogue turns using graph structures
• Improving LLM response consistency through relation graph prompting

Innovation

• Models multi-turn dialogues as directed relation graphs
• Uses action-triggered mechanisms to extract semantic relations
• Converts graph information into natural language prompts