What Models Know, How Well They Know It: Knowledge-Weighted Fine-Tuning for Learning When to Say "I Don't Know"

What Models Know, How Well They Know It: Knowledge-Weighted Fine-Tuning for Learning When to Say "I Don't Know" explores This method uses knowledge-weighted fine-tuning to enable LLMs to express uncertainty and avoid hallucinations for out-of-scope queries.. Commercial viability score: 5/10 in LLM Uncertainty.