SPRINT: One Forward Pass Generates Item Semantic IDs for Recommendation
_reachsumit · x · 2026-09-29
The paper SPRINT pursues single-step generative recommendation. Existing Semantic ID (SID) paradigms pay token by token: autoregressive models decode left-to-right while non-autoregressive ones still need multiple refinement rounds — both spend multiple forward passes per item, prohibitive for latency-sensitive recommenders.
Using a new perspective called average probability velocity, the authors prove it is fully determined by the average generation probability of each token, then parameterize and learn all token probabilities in a single forward pass with a bidirectional Transformer. A dual-level flow contrastive objective restores coherence among an item's tokens lost to independent generation.
More from Research
- New OCR leaderboard tests 14 open models on 2,165 historical pages; best hits 97.6% accuracy under $2 per 1k pages — IgorCarron · 2026-10-24
- Wavestone reverse-engineers 11 coding agents, maps the same 7-part harness inside all of them — blaizedsouza · 2026-10-24
- Looped Transformers heat up: 10 new papers on recurrent loop architectures — TheTuringPost · 2026-10-24
- FormalizeSampling: Sampling and diffusion convergence theorems machine-checked in Lean 4 — TianbaoX · 2026-10-24
- Alison Gopnik: What Kids Are 'Bad' At May Be Features, Not Flaws — AlisonGopnik · 2026-10-24
- Explicit Diophantine equation provably equivalent to Con(ZFC) constructed, Lean proof in progress — IgorCarron · 2026-10-24