Knowledge Distillation across Item and Code Spaces for Parallel Generative Recommendation
_reachsumit · x · 2026-10-26
A new arXiv paper proposes two complementary knowledge-distillation objectives for parallel semantic-ID generative recommenders. Existing semantic-ID models train on hard labels only, lacking distributional knowledge about alternative items; a sequential recommender can supply it, but teacher and student operate in item vs. code spaces. Item Distribution Matching (IDM) reconstructs a catalogue-level item distribution from the student's parallel code predictions and aligns it with the teacher's; Code Marginal Matching (CMM) projects the teacher's item probabilities onto each code position via fixed item-to-code assignments. Both transfer soft teacher knowledge while preserving parallel decoding.
More from Research
- HKU/ZJU/Fudan with Lightspeed launch HiSTEVE: hierarchical VLA agent playing Minecraft via native mouse and keyboard — tencent · 2026-10-26
- Sony researchers apply flow matching to four-stem stereo music source separation in FLOMSS — Shuyang Cui · 2026-10-26
- Zhejiang University and Tencent propose Easy-3Dfy: decoupling object generation and layout for single-image 3D scene construction — Ziang Zhang · 2026-10-26
- One week: 80-year-old cells reset to under 20, AI designing working proteins — rand_longevity · 2026-10-26
- Anthropic interpretability demo shows neurons combining into tiny pirate/wizard personas — Sauers_ · 2026-10-26
- 17-page HMM chapter PDF from Jurafsky & Martin's NLP textbook shared — burny_tech · 2026-10-26