LEGIT:基于标识符 Trie 的实时资格门控用于异构页面约束解码
原标题:LEGIT: Live Eligibility Gating over Identifier Tries for Constrained Decoding of Heterogeneous Pages
AI 摘要
该论文提出 LEGIT 约束,用于 Uber Eats 搜索生成式推荐系统在单次解码中生成整页结果(店铺及菜单项)时的合法性控制。LEGIT 通过深度优先编号的 trie 将实时资格过滤与去重规则转化为范围查询,并比较了 36 种实现。在单张 NVIDIA B200 GPU 上,最快的 CPU 遍历 trie 实现比直接 GPU 实现每 token 耗时降低 65.9%,约为无约束解码的 1.7 倍。
正文节选
marginparsep has been altered. topmargin has been altered. marginparwidth has been altered. marginparpush has been altered. The page layout violates the ICML style. Please do not change the page layout, or include packages like geometry, savetrees, or fullpage, which change it for you. We’re not able to reliably undo arbitrary changes to the style. Please remove the offending package(s), or layout-changing commands and try again. LEGIT: Live Eligibility Gating over Identifier Tries for Constrain