{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,3,18]],"date-time":"2026-03-18T02:38:51Z","timestamp":1773801531653,"version":"3.50.1"},"reference-count":0,"publisher":"Association for the Advancement of Artificial Intelligence (AAAI)","issue":"10","content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["AAAI"],"abstract":"<jats:p>Text-to-Video (T2V) generation has advanced greatly, yet maintaining consistency remains challenging, especially for tuning-free long video generation. \nWe attribute the consistency problem to cumulative deviations for long video generation at three levels: \nthe random noise lacking correlation results initial deviation between frames; \ndiscrepancy in semantic feature tokens between denoising network blocks gradually accumulates as the frame count grows, leading to greater deviations;\nattention mechanisms struggle to capture global relationships across distant frames in long videos. \nTo address these, we propose FreeMem, a tuning-free framework leveraging hierarchical memory update and injection: \nthe noise memory stabilizes consistency by manipulating low and high frequency components in the initial noise space; \nthe token memory combats inconsistency through adaptive fusion of historical and current semantic feature tokens between denoising network blocks; \nand the attention memory establishes persistent cache to model long-range relationships within self attention layers. \nEvaluated on VBench, FreeMem improves subject and background consistency matrics across various methods, offering a practical solution for low-cost, high-consistency long video generation.<\/jats:p>","DOI":"10.1609\/aaai.v40i10.37783","type":"journal-article","created":{"date-parts":[[2026,3,17]],"date-time":"2026-03-17T23:39:58Z","timestamp":1773790798000},"page":"8340-8348","source":"Crossref","is-referenced-by-count":0,"title":["FreeMem: Enhancing Consistency in Long Video Generation via Tuning-Free Memory"],"prefix":"10.1609","volume":"40","author":[{"given":"Jibin","family":"Peng","sequence":"first","affiliation":[]},{"given":"Di","family":"Lin","sequence":"additional","affiliation":[]},{"given":"Zhecheng","family":"Xu","sequence":"additional","affiliation":[]},{"given":"Haoran","family":"Lu","sequence":"additional","affiliation":[]},{"given":"Ruonan","family":"Liu","sequence":"additional","affiliation":[]},{"given":"Wuyuan","family":"Xie","sequence":"additional","affiliation":[]},{"given":"Miaohui","family":"Wang","sequence":"additional","affiliation":[]},{"given":"Lingyu","family":"Liang","sequence":"additional","affiliation":[]},{"given":"Yi","family":"Wang","sequence":"additional","affiliation":[]},{"given":"Qing","family":"Guo","sequence":"additional","affiliation":[]}],"member":"9382","published-online":{"date-parts":[[2026,3,14]]},"container-title":["Proceedings of the AAAI Conference on Artificial Intelligence"],"original-title":[],"link":[{"URL":"https:\/\/ojs.aaai.org\/index.php\/AAAI\/article\/download\/37783\/41745","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/ojs.aaai.org\/index.php\/AAAI\/article\/download\/37783\/41745","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,3,17]],"date-time":"2026-03-17T23:39:59Z","timestamp":1773790799000},"score":1,"resource":{"primary":{"URL":"https:\/\/ojs.aaai.org\/index.php\/AAAI\/article\/view\/37783"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,3,14]]},"references-count":0,"journal-issue":{"issue":"10","published-online":{"date-parts":[[2026,3,17]]}},"URL":"https:\/\/doi.org\/10.1609\/aaai.v40i10.37783","relation":{},"ISSN":["2374-3468","2159-5399"],"issn-type":[{"value":"2374-3468","type":"electronic"},{"value":"2159-5399","type":"print"}],"subject":[],"published":{"date-parts":[[2026,3,14]]}}}