Generation Constraint Scaling Can Mitigate Hallucination
This addresses hallucination issues in LLMs for users needing reliable text generation, though it is incremental as it builds on existing memory mechanisms.
The paper tackled hallucination in memory-augmented large language models by scaling a readout vector to constrain generation, achieving training-free mitigation and outperforming a state-of-the-art editing method in generation quality and runtime complexity for Wikipedia-like biography entries.
Addressing the issue of hallucinations in large language models (LLMs) is a critical challenge. As the cognitive mechanisms of hallucination have been related to memory, here we explore hallucination for LLM that is enabled with explicit memory mechanisms. We empirically demonstrate that by simply scaling the readout vector that constrains generation in a memory-augmented LLM decoder, hallucination mitigation can be achieved in a training-free manner. Our method is geometry-inspired and outperforms a state-of-the-art LLM editing method on the task of generation of Wikipedia-like biography entries both in terms of generation quality and runtime complexity.