Adaptive Semantic Capacity Allocation for Parallel Generative Recommendation

2026-08-10

AI Summary

Existing parallel generative recommendation methods rely on fixed and homogeneous semantic ID structures, which fail to accommodate the heterogeneous capacity demands across different semantic subspaces, leading to inefficient allocation of predictive capacity. This work proposes InforID, a novel framework that introduces, for the first time, a semantic-capacity-adaptive allocation mechanism. Under a global capacity budget, InforID jointly optimizes the ID length and codebook size for each semantic slot, thereby overcoming the limitations of conventional fixed-structure approaches. Built upon a parallel generative architecture, the method integrates information-driven semantic slot selection with dynamic codebook allocation, significantly improving recommendation accuracy while preserving single-step prediction efficiency. Empirical results validate the effectiveness and superiority of heterogeneous semantic ID structures enabled by this approach.

Abstract

Autoregressive semantic ID recommenders are constrained by expensive beam-search decoding, which limits the practical length of item identifiers. Parallel generation methods alleviate this bottleneck by predicting all semantic ID tokens simultaneously, enabling longer IDs. However, existing semantic ID methods still rely on manually predefined and homogeneous ID structures, where both the number of semantic slots and the codebook size of each slot are treated as fixed hyperparameters. This ignores the heterogeneous capacity demands of different semantic subspaces and may allocate prediction capacity to slots with limited utility. We show that uniformly expanding semantic slots can provide limited gains, indicating redundant capacity in homogeneous semantic IDs. We propose InforID, a lightweight adaptive semantic target construction framework for parallel generative recommendation. InforID allocates a fixed capacity budget across candidate semantic slots, thereby jointly determining the effective ID length and slot-specific codebook sizes. Experiments demonstrate improved recommendation accuracy under comparable capacity budgets while preserving one-step parallel prediction.

Problem

• semantic ID
• capacity allocation
• parallel generative recommendation
• heterogeneous subspaces
• codebook size

Innovation

• adaptive semantic allocation
• parallel generative recommendation
• semantic ID
• capacity budgeting
• InforID