Pre-training Limited Memory Language Models with Internal and External Knowledge
This addresses the issue of unreliable and uneditable knowledge in language models for users needing verifiable AI systems, though it is incremental in combining external knowledge with pre-training.
The paper tackles the problem of opaque factual knowledge encoding in neural language models by introducing Limited Memory Language Models (LMLM), which externalize factual knowledge to an external database during pre-training, achieving competitive performance compared to larger LLMs on standard benchmarks.
Neural language models are black-boxes--both linguistic patterns and factual knowledge are distributed across billions of opaque parameters. This entangled encoding makes it difficult to reliably inspect, verify, or update specific facts. We introduce Limited Memory Language Models (LMLM), a new class of language models that externalizes factual knowledge to external database during pre-training rather than memorizing them. Our pre-training approach strategically masks externally retrieved factual values from the training loss, thereby teaching the model to perform targeted lookups rather than relying on memorization in model weights. Our experiments demonstrate that LMLMs achieve competitive performance compared to significantly larger LLMs on standard benchmarks, while offering the advantages of explicit, editable, and verifiable knowledge bases.