Metaverse,　as　a　revolutionary　technology　that　changes　the　way　of　human　interaction,　brings　new　challenges　to　content　delivery　services　due　to　the　extensive　data　transmission　and　personalized　service　requirements.　To　ensure　a　personalized　user　experiences　while　improving　the　utilization　of　heterogeneous　network　resources,　a　user-centric　many-objective　metaverse　content　delivery　framework　is　proposed　to　optimize　content　delivery　through　user　attention　awareness.　This　framework　addresses　two　key　subproblems　in　metaverse　content　delivery　by　investigating　user-centric　many-objective　cooperative　content　caching　and　deep　reinforcement　learning　(DRL)-based　request　routing.　The　user-centric　many-objective　cooperative　content　caching　is　proposed　to　dynamically　combine　three　basic　preference　prediction　results　to　predict　user　preferences　and　control　network　resource　allocation,　which　can　simultaneously　optimize　prediction　precision,　delay,　offloaded　traffic,　and　load　balancing.　In　DRL-based　request　routing,　the　reward　function　is　designed　to　enable　the　optimization　of　multiple　objectives.　The　multi-objective　DRL　routing　algorithm　is　employed　to　continuously　observe　network　states　and　make　adaptive　routing　decisions　in　response　to　user　requests.　In　the　simulation,　a　movie　dataset　is　employed　to　simulate　user　requests　and　support　user　attention　awareness.　The　results　show　that　the　proposed　content　delivery　framework　outperforms　existing　basic　prediction　algorithms　and　other　content　delivery　algorithms　on　four　evaluation　indicators.　©　2013　IEEE.