CLLGOct 10, 2023

Teaching Language Models to Hallucinate Less with Synthetic Tasks

arXiv:2310.06827v344 citationsh-index: 26
Originality Incremental advance
AI Analysis

This addresses the challenge of mitigating hallucinations in LLMs for applications like document-based QA and clinical report generation, though it is incremental as it builds on existing prefix-tuning methods.

The paper tackles the problem of reducing hallucinations in large language models on abstractive summarization tasks by optimizing system messages via prefix-tuning on a synthetic task, and shows that this reduces hallucination across three realistic tasks for two 13B-parameter models.

Large language models (LLMs) frequently hallucinate on abstractive summarization tasks such as document-based question-answering, meeting summarization, and clinical report generation, even though all necessary information is included in context. However, optimizing LLMs to hallucinate less on these tasks is challenging, as hallucination is hard to efficiently evaluate at each optimization step. In this work, we show that reducing hallucination on a synthetic task can also reduce hallucination on real-world downstream tasks. Our method, SynTra, first designs a synthetic task where hallucinations are easy to elicit and measure. It next optimizes the LLM's system message via prefix-tuning on the synthetic task, and finally transfers the system message to realistic, hard-to-optimize tasks. Across three realistic abstractive summarization tasks, SynTra reduces hallucination for two 13B-parameter LLMs using only a synthetic retrieval task for supervision. We also find that optimizing the system message rather than the model weights can be critical; fine-tuning the entire model on the synthetic task can counterintuitively increase hallucination. Overall, SynTra demonstrates that the extra flexibility of working with synthetic data can help mitigate undesired behaviors in practice.

Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes