CapsID uses probabilistic capsule routing and confidence-based termination to generate variable-length semantic IDs, improving recall by 9.6% over strong baselines with half the latency of dual-representation systems.
Rethinking generative recommender to- kenizer: Recsys-native encoding and semantic quantization beyond llms
2 Pith papers cite this work. Polarity classification is still indexing.
2
Pith papers citing it
fields
cs.IR 2years
2026 2verdicts
UNVERDICTED 2representative citing papers
DIG unifies ranking and retrieval by training the tokenizer jointly inside a ranking model, producing improved models for both from a single run.
citing papers explorer
-
CapsID: Soft-Routed Variable-Length Semantic IDs for Generative Recommendation
CapsID uses probabilistic capsule routing and confidence-based termination to generate variable-length semantic IDs, improving recall by 9.6% over strong baselines with half the latency of dual-representation systems.
-
Discrimination Is Generation: Unifying Ranking and Retrieval from a Tokenizer Perspective
DIG unifies ranking and retrieval by training the tokenizer jointly inside a ranking model, producing improved models for both from a single run.