KoCo: Conditioning Language Model Pre-training on Knowledge Coordinates

TL;DR

KoCo introduces a method to incorporate explicit real-world knowledge context into language model pre-training by mapping documents into semantic coordinates, improving downstream task performance and reducing hallucinations.

Contribution

The paper proposes Knowledge Coordinate Conditioning (KoCo), a novel approach that enhances language models with explicit knowledge context during pre-training.

Findings

KoCo improves performance across 10 downstream tasks.

Pre-training convergence accelerates by approximately 30%.

Model better distinguishes facts from noise, reducing hallucinations.

Abstract

Standard Large Language Model (LLM) pre-training typically treats corpora as flattened token sequences, often overlooking the real-world context that humans naturally rely on to contextualize information. To bridge this gap, we introduce Knowledge Coordinate Conditioning (KoCo), a simple method that maps every document into a three-dimensional semantic coordinate. By prepending these coordinates as textual prefixes for pre-training, we aim to equip the model with explicit contextual awareness to learn the documents within the real-world knowledge structure. Experiment results demonstrate that KoCo significantly enhances performance across 10 downstream tasks and accelerates pre-training convergence by approximately 30\%. Furthermore, our analysis indicates that explicitly modeling knowledge coordinates helps the model distinguish stable facts from noise, effectively mitigating hallucination in generated outputs.