Tag
This paper introduces MORSE, a compression-aware method for evidence-preserving context ordering in large language models that improves evidence retention and downstream QA performance.
An AI tool has been developed that can convert any paper into an agent capable of collaborating and answering complex queries.
This paper studies the complexity of answering two-way (conjunctive) regular path queries over knowledge bases with linear and guarded existential rules, establishing completeness results for data and combined complexity.
This paper studies Controlled Query Evaluation (CQE) for Description Logic ontologies with epistemic dependency policies, introducing a new semantics based on minimal policy violation that achieves polynomial-time query answering for DL-Lite ontologies.