Abstract:As Large Language Models (LLMs) continue to advance in their ability to write human-like text, a key challenge remains around their tendency to hallucinate generating content that appears factual but is ungrounded. This issue of hallucination is arguably the biggest hindrance to safely deploying these powerful LLMs into real-world production systems that impact people's lives. The journey toward widespread adoption of LLMs in practical settings heavily relies on addressing and mitigating hallucinations. Unlike traditional AI systems focused on limited tasks, LLMs have been exposed to vast amounts of online text data during training. While this allows them to display impressive language fluency, it also means they are capable of extrapolating information from the biases in training data, misinterpreting ambiguous prompts, or modifying the information to align superficially with the input. This becomes hugely alarming when we rely on language generation capabilities for sensitive applications, such as summarizing medical records, financial analysis reports, etc. This paper presents a comprehensive survey of over 32 techniques developed to mitigate hallucination in LLMs. Notable among these are Retrieval Augmented Generation (Lewis et al, 2021), Knowledge Retrieval (Varshney et al,2023), CoNLI (Lei et al, 2023), and CoVe (Dhuliawala et al, 2023). Furthermore, we introduce a detailed taxonomy categorizing these methods based on various parameters, such as dataset utilization, common tasks, feedback mechanisms, and retriever types. This classification helps distinguish the diverse approaches specifically designed to tackle hallucination issues in LLMs. Additionally, we analyze the challenges and limitations inherent in these techniques, providing a solid foundation for future research in addressing hallucinations and related phenomena within the realm of LLMs.

Detecting and Mitigating the Ungrounded Hallucinations in Text Generation by LLMs

Mitigating Entity-Level Hallucination in Large Language Models

Developing a Reliable, General-Purpose Hallucination Detection and Mitigation Service: Insights and Lessons Learned

Detecting and Mitigating Hallucination in Large Vision Language Models via Fine-Grained AI Feedback

Chain of Natural Language Inference for Reducing Large Language Model Ungrounded Hallucinations

A Stitch in Time Saves Nine: Detecting and Mitigating Hallucinations of LLMs by Validating Low-Confidence Generation

Hallucination Detection and Hallucination Mitigation: An Investigation

Towards Mitigating Hallucination in Large Language Models via Self-Reflection

A Comprehensive Survey of Hallucination Mitigation Techniques in Large Language Models

Insights into Classifying and Mitigating LLMs' Hallucinations

Retrieve Only When It Needs: Adaptive Retrieval Augmentation for Hallucination Mitigation in Large Language Models

Mechanistic Understanding and Mitigation of Language Model Non-Factual Hallucinations

Hallucination Detection: Robustly Discerning Reliable Answers in Large Language Models

Exploring and Evaluating Hallucinations in LLM-Powered Code Generation

Towards Detecting LLMs Hallucination via Markov Chain-based Multi-agent Debate Framework

A Survey on Hallucination in Large Language Models: Principles, Taxonomy, Challenges, and Open Questions

Zero-Resource Hallucination Prevention for Large Language Models

The Dawn After the Dark: An Empirical Study on Factuality Hallucination in Large Language Models

Analyzing and Mitigating Object Hallucination in Large Vision-Language Models

Look Within, Why LLMs Hallucinate: A Causal Perspective