IDEA

Large language models have shown unprecedented abilities in generating linguistically coherent and syntactically correct natural language output. However, they often return incorrect and inconsistent answers to input questions. Due to the complexity and uninterpretability of the internally learned representations, it is challenging to modify language models such that they provide correct and consistent results. The data management community has developed various methods and tools for providing consistent answers over inconsistent datasets. In these methods, users specify the desired properties of data in a domain in the form of high-level declarative constraints. This approach has provided usable and scalable methods to delivering consistent information from inconsistent datasets. We propose to build upon this success and leverage these methods to modify language models such that they deliver consistent and accurate results.

Publications

• Towards Consistent Language Models Using Controlled Prompting and Decoding  [Slides]  [Poster]  [BibTeX]
  Jasmin Mousavi and Arash Termehchy
  AAAI workshop on Neuro-Symbolic Learning and Reasoning in the Era of Large Language Models, February, 2024.
• Towards Consistent Language Models Using Declarative Constraints  [BibTeX]
  Jasmin Mousavi and Arash Termehchy
  Technical Report (in arXiv): arXiv:2312.15472 [cs.DB], December 2023
• Towards Consistent Large Language Models Using Declarative Constraints  [Slides]  [BibTeX]
  Jasmin Mousavi and Arash Termehchy
  The Proceedings of VLDB Workshop on Databases and Large Language Models (LLMDB), August, 2023.

People

• Arash Termehchy
• Jasmin Mousavi