The　title　of　an　academic　paper　encapsulates　its　core　knowledge　in　a　succinct　manner,　while　the　abstract　provides　a　brief　yet　comprehensive　summary　of　the　work.　Crafting　an　effective　title　is　essential　for　enhancing　both　the　dissemination　and　retrieval　of　scholarly　research.　To　address　this　need,　we　propose　an　automatic　title　generation　model　based　on　abstractive　summarization　techniques,　aimed　at　assisting　researchers　in　the　title　drafting　process.　This　study　introduces　a　novel　model　for　generating　titles　for　Chinese　academic　papers,　utilizing　the　Transformer　architecture　in　conjunction　with　the　pre-trained　BERT　model.　The　challenges　inherent　in　Chinese　word　segmentation　are　explored,　and　the　effects　of　character-based　versus　word-based　tokenization　on　model　performance　are　thoroughly　examined.　Furthermore,　a　robust　evaluation　framework　incorporating　both　quantitative　metrics　and　human　evaluations　is　implemented　to　assess　the　effectiveness　of　the　proposed　approach.　Experimental　results　demonstrate　that　the　Transformer　model,　augmented　by　pre-training,　outperforms　established　baselines,　achieving　a　Rouge-L　score　of　0.441.　Further　analysis　reveals　that　the　generated　titles　accurately　capture　the　essential　content　of　the　papers,　thus　facilitating　decision-making　in　academic　title　generation.　This　method　shows　considerable　potential　for　improving　the　efficiency　and　quality　of　academic　writing　in　the　Chinese　language.　©　China　Computer　Federation　(CCF)　2025.