Bactrainus: Optimizing Large Language Models for Multi-hop Complex Question Answering Tasks
Researchers have developed a framework called Bactrainus to improve large language models' ability to answer complex questions that require multiple steps. The model separates tasks into paragraph selection, supporting sentence identification, and answer generation. Experiments showed that using the full candidate context instead of just gold supporting facts improved performance by 17-21 points. However, the largest gains came from adapting readers and controlling evidence a
Researchers have developed a framework called Bactrainus to improve large language models' ability to answer complex questions that require multiple steps. The model separates tasks into paragraph selection, supporting sentence identification, and answer generation. Experiments showed that using the full candidate context instead of just gold supporting facts improved performance by 17-21 points. However, the largest gains came from adapting readers and controlling evidence at the sentence level.
---
Why it matters: This matters to researchers in AI because it shows how to improve the performance of large language models on complex question answering tasks, which is a crucial application area for natural language processing.
Source: https://arxiv.org/abs/2501.06286
This article was originally published at: https://arxiv.org/abs/2501.06286