Search papers, labs, and topics across Lattice.
This paper introduces DIRECTOR, a novel framework for dynamic index-based recommendation that addresses the limitations of both autoregressive and non-autoregressive reranking methods. By employing transport-optimized retrieval and mapping candidate items into a continuous latent space, DIRECTOR enables parallel processing while maintaining global structural coordination, thus avoiding issues like duplicate selections. Extensive experiments show that DIRECTOR significantly outperforms existing reranking baselines in large-scale industrial settings, highlighting its practical efficacy in recommendation systems.
Achieving duplicate-free item selection in recommendation systems without sacrificing efficiency, DIRECTOR revolutionizes how we approach reranking by leveraging transport-optimized retrieval.
Reranking is a combinatorial decision problem that aims to select and order a high-utility slate from a request-specific candidate set. A major line of generative rerankers adopts autoregressive (AR) models, which construct the slate one position at a time to capture inter-position dependencies. However, under practical greedy or bounded-width decoding, prefix-based search may prematurely prune globally promising permutations and incurs inherently sequential latency, restricting the effective search space under a fixed serving budget. Non-autoregressive (NAR) alternatives alleviate this efficiency bottleneck through position-parallel prediction, but naive position-wise factorization treats different positions too independently, leading to insufficient cross-position coordination and potentially duplicate or conflicting item selections. To retain parallel efficiency while introducing global structural coordination, we propose Dynamic Index-based RECommendation with Transport-Optimized Retrieval (DIRECTOR), a transport-guided parallel reranking framework. DIRECTOR maps candidate items into a continuous latent space and generates request-conditioned dynamic retrieval indices for all target positions in parallel. During training, it uses entropy-regularized OT to provide conflict-aware supervision; at inference, it directly performs global hard matching on similarity matrix, producing duplicate-free slates without iterative transport. To further align the generator with an opaque list-wise evaluator that returns only a scalar utility, we introduce a prefix-anchored credit assignment mechanism that converts the global reward into position-specific training signals. Extensive offline and online experiments demonstrate that DIRECTOR consistently outperforms strong reranking baselines, achieving significant improvement in large-scale industrial recommendation scenarios.
