WikiChat: Stopping the Hallucination of Large Language Model Chatbots by Few-Shot Grounding on Wikipedia

Sina Semnani, Violet Yao, Heidi Chenyu Zhang, Monica Lam

Submission Type: Regular Long Paper

Submission Track: Dialogue and Interactive Systems

Keywords: large language models, hallucination, knowledge-grounded dialogue

TL;DR: We build and evaluate a fully few-shot LLM-based chatbot that almost never hallucinates, then distill it into a 7B-parameter LLaMA model.

Abstract:

This paper presents the first few-shot LLM-based chatbot that almost never hallucinates and has high conversationality and low latency. WikiChat is grounded on the English Wikipedia, the largest curated free-text corpus.

WikiChat generates a response from an LLM, retains only the grounded facts, and combines them with additional information it retrieves
from the corpus to form factual and engaging responses. We distill WikiChat based on GPT-4 into a 7B-parameter LLaMA model with minimal loss of quality, to significantly improve its latency, cost and privacy, and facilitate research and deployment.

Using a novel hybrid human-and-LLM evaluation methodology, we show that our best system achieves 97.3% factual accuracy in simulated
conversations. It significantly outperforms all retrieval-based and LLM-based baselines, and by 3.9%, 38.6% and 51.0% on head, tail and recent knowledge compared to GPT-4. Compared to previous state-of-the-art retrieval-based chatbots, WikiChat is also significantly more informative and engaging, just like an LLM.

WikiChat achieves 97.9% factual accuracy in conversations with human users about recent topics, 55.0% better than GPT-4, while receiving significantly higher user ratings and more favorable comments.

Submission Number: 2085