AI RESEARCH

Make It Long, Keep It Fast: End-to-End 10k-Sequence Modeling at Billion Scale on Douyin Recommendation

arXiv CS.LG

ArXi:2511.06077v2 Announce Type: replace Short-video recommenders such as Douyin must exploit extremely long user histories without breaking latency or cost budgets. We present an end-to-end system that scales long-sequence modeling to 10k-length histories in production. First, we