Papers by Haein Jung
SELF-EXPERTISE: Knowledge-based Instruction Dataset Augmentation for a Legal Expert Language Model (2024.findings-naacl)
Copied to clipboard
| Challenge: | generating instructions and outputs from LLMs can produce unintentionally inaccurate or misleading information. |
| Approach: | They propose to generate an instruction dataset in the legal domain from a seed dataset by extracting knowledge from the outputs of the seed dataset. |
| Outcome: | The proposed method reduces hallucinations in automatic instruction dataset augmentation. |