Co-Evolving Structured Knowledge and Reasoning in Language Models
Abstract
KBevo jointly learns to build and reason over structured knowledge bases for question answering, improving reasoning and controllability through end-to-end optimization.
Retrieval-augmented methods improve factual accuracy by grounding language models in external knowledge, but retrieving over unstructured text often introduces irrelevant context and offers limited control over the retrieved information. Structured knowledge bases offer a more controllable alternative, yet they are expensive to construct and often brittle to reason over. To address these limitations, we propose KBevo: a co-evolving framework that jointly learns to construct a structured knowledge base and reason over it for knowledge-intensive question answering. By optimizing both components end-to-end with QA outcome rewards, our method enables reasoning success to directly improve the quality of the constructed knowledge base. This leads to larger, better-connected knowledge structures with higher answer reachability, while also improving compositional factual reasoning and controllability compared to standard retrieval baselines.
Get this paper in your agent:
hf papers read 2608.26386 Don't have the latest CLI?
curl -LsSf https://hf.co/cli/install.sh | bash Models citing this paper 4
kilian-group/KBevo-Qwen3-4B-SFT
Datasets citing this paper 1
kilian-group/KBevo-SFT-hotpotqa-6k
Spaces citing this paper 0
No Space linking this paper
Collections including this paper 0
No Collection including this paper