GraphReader: Building Graph-based Agent to Enhance Long-Context Abilities of Large Language Models

TL;DR

GraphReader introduces a graph-based agent system that structures long texts into graphs and autonomously explores them, significantly enhancing large language models' ability to process and reason over very long contexts.

Contribution

The paper presents a novel graph-based agent system that improves long-context understanding in LLMs by autonomous graph exploration and rational planning.

Findings

Outperforms GPT-4-128k on LV-Eval dataset with 4k context window.

Achieves superior results on four challenging benchmarks.

Effectively handles context lengths from 16k to 256k.

Abstract

Long-context capabilities are essential for large language models (LLMs) to tackle complex and long-input tasks. Despite numerous efforts made to optimize LLMs for long contexts, challenges persist in robustly processing long inputs. In this paper, we introduce GraphReader, a graph-based agent system designed to handle long texts by structuring them into a graph and employing an agent to explore this graph autonomously. Upon receiving a question, the agent first undertakes a step-by-step analysis and devises a rational plan. It then invokes a set of predefined functions to read node content and neighbors, facilitating a coarse-to-fine exploration of the graph. Throughout the exploration, the agent continuously records new insights and reflects on current circumstances to optimize the process until it has gathered sufficient information to generate an answer. Experimental results on the LV-Eval dataset reveal that GraphReader, using a 4k context window, consistently outperforms GPT-4-128k across context lengths from 16k to 256k by a large margin. Additionally, our approach demonstrates superior performance on four challenging single-hop and multi-hop benchmarks.