Abstract: [Purpose/Significance] Oriented toward the construction of an independent knowledge system for philosophy and social sciences with Chinese characteristics, this paper explores an LLM-driven framework for innovation evaluation, aiming to shift evaluation toward content understanding and evidence-based assessment. [Design/Methodology] This paper constructs a LLM-driven innovation evaluation framework for philosophy and social sciences, proposing a five-dimensional evaluation system—value guidance, problem orientation, innovation contribution, argumentative support, and outcome impact—operationalized into nineteen indicators. Based on this, it designs an implementation pathway using full-text data and multi-source information, supported by model-based analysis and expert involvement. [Findings/Conclusion] LLMs hold significant promise for enhancing content recognition and evidence-based support in the evaluation of innovation, but currently function mainly as auxiliary evaluation tool. [Originality/Value] Grounded in the intrinsic requisites of academic innovation assessment for philosophy and social sciences, this paper constructs an LLM-driven evaluation indicator system and its implementation pathway, providing a reference for reforming innovation evaluation and improving the evaluation system of philosophy and social sciences in China.

Keywords: Philosophy and social sciences, Innovation evaluation, Large language models, Paradigm transformation, Independent knowledge system