CapsID: Soft-Routed Variable-Length Semantic IDs for Generative Recommendation

TL;DR

CapsID introduces a capsule routing-based tokenizer for generative recommendation, improving semantic representation and efficiency over traditional hard quantization methods.

Contribution

It replaces hard residual quantization with capsule routing, enhancing semantic fidelity and reducing inference latency in generative recommendation systems.

Findings

CapsID improves Recall@10 by 9.6% on average over ReSID.

CapsID matches or exceeds COBRA-style systems on benchmarks.

CapsID runs at 51% of the inference latency of comparable systems.

Abstract

Generative recommendation maps each item to a sequence of Semantic IDs (SIDs) and recasts retrieval as autoregressive token generation. In this paradigm the main bottleneck is the tokenizer rather than the Transformer: residual vector quantization with a hard nearest-neighbor assignment at every layer collapses multi-faceted item semantics at cluster boundaries and propagates early errors to later SID positions. A common workaround is to append a dense vector or attribute prefix to the SID, but this dual-representation design inflates inference cost and gives up the simplicity of a generative interface. We address the bottleneck at the tokenizer itself. CAPSID replaces hard residual quantization with capsule routing: at each layer an item probabilistically routes to several semantic capsules, the residual is updated by the routed reconstruction rather than by a single winning code, and the SID terminates once the active capsule's confidence is high enough. On top of CAPSID, SEMANTICBPE composes adjacent SID tokens into reusable subwords by combining their co-occurrence with their embedding compatibility. On Amazon Beauty, Sports, Toys, and a 35M-item proprietary industrial catalog, CAPSID+SEMANTICBPE improves Recall at 10 by 9.6% on average over ReSID, the strongest single-representation baseline, and matches or exceeds a COBRA-style sparse-dense system on every public benchmark while running at 51% of its inference latency. Ablations show that soft routing, iterative agreement, and confidence-driven length each contribute independently, and the gains are largest on tail items where boundary semantics dominate.