Omne-R1: Learning to Reason with Memory for Multi-hop Question Answering
This addresses the problem of limited knowledge graphs and QA data for multi-hop reasoning, though it appears incremental in its approach.
The paper tackles multi-hop question answering on schema-free knowledge graphs by introducing Omne-R1, a method using multi-stage training with reinforcement learning and fine-tuning, which shows significant improvements, especially on complex 3+ hop questions.
This paper introduces Omne-R1, a novel approach designed to enhance multi-hop question answering capabilities on schema-free knowledge graphs by integrating advanced reasoning models. Our method employs a multi-stage training workflow, including two reinforcement learning phases and one supervised fine-tuning phase. We address the challenge of limited suitable knowledge graphs and QA data by constructing domain-independent knowledge graphs and auto-generating QA pairs. Experimental results show significant improvements in answering multi-hop questions, with notable performance gains on more complex 3+ hop questions. Our proposed training framework demonstrates strong generalization abilities across diverse knowledge domains.