Curriculum learning in natural language generation

• Applicant: Professorin Dr. Milica Gasic
• Subject Area: Image and Language Processing, Computer Graphics and Visualisation, Human Computer Interaction, Ubiquitous and Wearable Computing
• Term: since 2024
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 548039008

Hallucination detection and mitigation are becoming increasingly important and urgent topics in natural language processing (NLP) field as large language models become used and relied upon by to millions of users worldwide. The goal of this project is to better understand the origins of hallucinations and eventually reduce their magnitude and severity in the output of NLG models. This project aims to address the issue of hallucinations by incorporating human-learning-inspired mechanisms into a curriculum learning. Central hypothesis of this project is that one possible source for hallucinations is that all data are learned at once, creating associations when the relevance of the input is not yet clear.

• DFG Programme: Research Grants