Fine-tuning LLMs on known versus unknown facts creates a factuality gap that in-context prompting can largely erase, according to experiments and a knowledge-graph model.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
From Parameters to Prompts: Understanding and Mitigating the Factuality Gap between Fine-Tuned LLMs
Fine-tuning LLMs on known versus unknown facts creates a factuality gap that in-context prompting can largely erase, according to experiments and a knowledge-graph model.