MAC-Tuning: LLM Multi-Compositional Problem Reasoning with Enhanced Knowledge Boundary Awareness
This addresses hallucination issues for LLM users in applications requiring simultaneous multi-question answering, though it builds incrementally on prior single-problem confidence estimation work.
The paper tackles the problem of LLM hallucination in multi-problem settings by proposing MAC-Tuning, a method that separates answer prediction and confidence estimation during fine-tuning, achieving up to 25% improvement in average precision over baselines.
The hallucination of non-existent facts by LLMs is an important problem given its widespread adoption across various applications. Previous research addresses this problem by analyzing the internal parameterized knowledge boundaries to estimate confidence. However, these studies focus on the single-problem setting and have not explored the more challenging multi-problem setting, which requires accurately answering multiple questions simultaneously. We introduce a novel method for the multi-problem setting, Multiple Answers and Confidence Stepwise Tuning (MAC-Tuning), that separates the learning of answer prediction and confidence estimation during fine-tuning on instruction data. Extensive experiments demonstrate that our method outperforms baselines by up to 25\% in average precision.