Collaborative decoding of critical tokens for boosting factuality of large language models
This work addresses the issue of factuality loss in fine-tuned LLMs, which is an incremental improvement for users needing more reliable AI-generated content.
The paper tackles the problem of hallucination in large language models after fine-tuning by introducing a collaborative decoding framework that uses a critical token classifier to decide between models for token generation, resulting in significant reduction of model hallucination.
The most common training pipeline for large language models includes pretraining, finetuning and aligning phases, with their respective resulting models, such as the pretrained model and the finetuned model. Finetuned and aligned models show improved abilities of instruction following and safe generation, however their abilities to stay factual about the world are impacted by the finetuning process. Furthermore, the common practice of using sampling during generation also increases chances of hallucination. In this work, we introduce a collaborative decoding framework to harness the high factuality within pretrained models through the concept of critical tokens. We first design a critical token classifier to decide which model to use for the next token, and subsequently generates the next token using different decoding strategies. Experiments with different models and datasets show that our decoding framework is able to reduce model hallucination significantly, showcasing the importance of the collaborative decoding framework.