What Algorithms Can Transformers Learn; Reasoning Agent for Graphs; Supervised Fine-Tuning; Context Understanding in LLMs; and More.

Editor's Paper Recommendations

Knowledge Editing for Large Language Models : A Survey: Large language models (LLMs) have recently transformed academic and industrial landscapes due to their remarkable capacity to understand, analyze, and generate texts based on their vast knowledge and reasoning ability. Nevertheless, one major drawback of LLMs is their substantial computational cost for pre-training due to their unprecedented amounts of parameters. The disadvantage is exacerbated when new knowledge frequently needs to be introduced into the pre-trained model. Therefore, developing effective and efficient techniques to update pre-trained LLMs is imperative. Traditional methods encode new knowledge in pre-trained LLMs through direct fine-tuning. However, naively re-training LLMs can be computationally intensive and risks degenerating valuable pre-trained knowledge irrelevant to the update in the model. Knowledge-based Model Editing (KME) has recently attracted increasing attention, aiming to precisely modify the LLMs to incorporate specific knowledge without negatively influencing irrelevant knowledge. In this survey, we aim to provide a comprehensive and in-depth overview of recent advances in the field of KME. We first introduce a general formulation of KME to encompass different KME strategies. Afterward, we provide an innovative taxonomy of KME techniques based on how the new knowledge is introduced into pre-trained LLMs. We investigate existing KME strategies while analyzing key insights, advantages, and limitations of methods from each category. Moreover, representative metrics, datasets, and applications of KME are introduced accordingly. Finally, we'd like to provide an in-depth analysis regarding the practicality and remaining challenges of KME and suggest promising research directions for further advancement in this field.

Graph Agent: Explicit Reasoning Agent for Graphs : Graph embedding methods such as Graph Neural Networks (GNNs) and Graph Transformers have contributed to developing graph reasoning algorithms for various tasks on knowledge graphs. However, the lack of interpretability and explainability of graph embedding methods has limited their applicability in scenarios requiring explicit reasoning. This paper introduces the Graph Agent (GA), an intelligent agent methodology of leveraging large language models (LLMs), inductive-deductive reasoning modules, and long-term memory for knowledge graph reasoning tasks. GA integrates aspects of symbolic reasoning and existing graph embedding methods to provide an innovative approach for complex graph reasoning tasks. By converting graph structures into textual data, GA enables LLMs to process, reason, and provide predictions alongside human-interpretable explanations. The effectiveness of the GA was evaluated on node classification and link prediction tasks. Results showed that GA reached state-of-the-art performance, demonstrating accuracy of 90.65%, 95.48%, and 89.32% on Cora, PubMed, and PrimeKG datasets, respectively. Compared to existing GNN and transformer models, GA offered the advantages of explicit reasoning ability, freedom of training, and easy adaptation to various graph reasoning tasks.

What Algorithms Can Transformers Learn? A Study in Length Generalization : Large language models exhibit surprising emergent generalization properties yet struggle with many simple reasoning tasks, such as arithmetic and parity. This raises the question of if and when Transformer models can learn the true algorithm for solving a task. We study the scope of Transformers' abilities in the specific setting of length generalization on algorithmic tasks. Here, we propose a unifying framework to understand when and how Transformers can exhibit strong length generalization on a given task. Specifically, we leverage RASP (Weiss et al., 2021), a programming language designed for the computational model of a transformer, and introduce the RASP-Generalization Conjecture: Transformers tend to generalize on a task if the task can be solved by a short RASP program that works for all input lengths. This simple conjecture remarkably captures the most known instances of length generalization on algorithmic tasks. Moreover, we leverage our insights to drastically improve generalization performance on traditionally hard tasks (such as parity and addition). Theoretically, we give a simple example where the "min-degree interpolator" model of learning from Abbe et al. (2023) does not correctly predict Transformers' out-of-distribution behavior, but our conjecture does. Overall, our work provides a novel perspective on the mechanisms of compositional generalization and the algorithmic capabilities of Transformers.

?

Meet SingleStore Pro Max, the Powerhouse Edition

In the rapidly changing landscape of AI and real-time analytics, the foundation of your applications—the data platform—is no longer an optional frill but a must-have. It's the springboard for innovation, the hidden force behind every breakthrough application.

Introducing SingleStore Pro Max: The Powerhouse Edition

Link

--

Are you looking to advertise a product, job opening, or event to an audience of over 40,000 AI researchers and engineers? Please reach out to us on?LinkedIn? to explore your options.

Enjoy the newsletter? Help us make it bigger and better by sharing it with colleagues and friends.

--

Industry Insights

• Supervised Fine-Tuning (SFT) with Large Language Models
• Temporal Graph Learning in 2024
• Preference Tuning LLMs with Direct Preference Optimization Methods
• A Groundbreaking AI Technique to Expand Context Understanding in Large Language Models

Growth Zone

• How to Motivate Employees to Go Beyond Their Jobs

?

Expert Advice