Grounding and Evaluation of Large Language Models: Practical Challenges and Lessons Learned (Survey)
Grounding and Evaluation for Large Language Models: Practical Challenges and Lessons Learned (Survey)
AI summary
A survey accompanying a KDD 2024 tutorial focuses on the grounding and evaluation of large language models and generative AI, noting that AI systems in high-risk domains need to evaluate responsible AI dimensions such as robustness, bias, safety, and interpretability in addition to accuracy and quality. The survey summarizes multiple categories of harms brought by generative AI, including model hallucination, harmful and manipulative content, and copyright infringement, and reviews the latest methods for addressing these harms and the remaining open challenges.