rLLM: Relational Table Learning with LLMs

2024-07-29

arXiv.org

AI Summary

To address the low efficiency and lack of a unified framework in relational table learning (RTL) model development, this paper introduces rLLM—a PyTorch-based open-source library enabling modular, collaborative modeling between large language models (LLMs) and graph/table neural networks (GNNs/TabNNs). We propose a novel “compose-align-co-train” RTL paradigm and a standardized module decomposition methodology, significantly enhancing rapid model construction and reproducibility. Concurrently, we release three high-quality, benchmark datasets: TML1M (million-scale), TLF2K (fine-grained semantics), and TACM12K (cross-domain multi-task). rLLM has been widely adopted by both academia and industry, establishing a scalable, user-friendly, and unified infrastructure for RTL research and development.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: (Large) Language Models
• Knowledge Representation and Reasoning: Knowledge Representation Languages

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
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation

Abstract

We introduce rLLM (relationLLM), a PyTorch library designed for Relational Table Learning (RTL) with Large Language Models (LLMs). The core idea is to decompose state-of-the-art Graph Neural Networks, LLMs, and Table Neural Networks into standardized modules, to enable the fast construction of novel RTL-type models in a simple"combine, align, and co-train"manner. To illustrate the usage of rLLM, we introduce a simple RTL method named extbf{BRIDGE}. Additionally, we present three novel relational tabular datasets (TML1M, TLF2K, and TACM12K) by enhancing classic datasets. We hope rLLM can serve as a useful and easy-to-use development framework for RTL-related tasks. Our code is available at: https://github.com/rllm-project/rllm.

Problem

• Develops a PyTorch library for relational table learning with LLMs
• Decomposes neural networks into modules for fast model construction
• Introduces novel datasets and methods for relational table tasks

Innovation

• Decomposes GNNs, LLMs, TNNs into modules
• Enables fast model construction via combining modules
• Introduces BRIDGE method for relational table learning