Research map / Label-efficient and robust learning / Recognition and long-tail
Inductive Generative Recommendation via Retrieval-based Speculation
Yijie Ding, Jiacheng Li, Julian McAuley, Yupeng Hou
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add inductive drafting to SpecGR++ (ablation) | SpecGR++ · Amazon Games | Recall@50: 0.0609 -> 0.0717 | Improved |
| add likelihood score adjustment (exclude item identification token) (ablation) | SpecGR++ · Amazon Games | Recall@50: 0.0712 -> 0.0717 | Improved |
| add guided re-drafting (ablation) | SpecGR++ · Amazon Games | Recall@50: 0.0611 -> 0.0717 | Improved |
| add verification-based re-ranking (ablation) | SpecGR++ · Amazon Games | Recall@50: 0.0703 -> 0.0717 | Improved |
| add adaptive exiting (ablation) | SpecGR++ · Amazon Office | Recall@10: 0.0203 -> 0.0250 | Improved |
| use dedicated contrastive pretraining for the GR encoder instead of raw TIGER encoder states (ablation) | SpecGR++ · Amazon Games | Recall@50: 0.0581 -> 0.0717 | Improved |
| add learning-to-rank fine-tuning after contrastive pretraining (ablation) | SpecGR++ · Amazon Games | Recall@50: 0.0692 -> 0.0717 | Improved |
| add SpecGR drafting and verification to a TIGER backbone | TIGER · Amazon Games unseen subset | Recall@50: TIGER fails on unseen (-) -> 0.0312 | Improved |
| increase draft size (ablation) | SpecGR++ · Amazon Games | Unseen and in-sample Recall@50 | Mixed |
| increase beam size (ablation) | SpecGR++ · Amazon Games | Unseen and in-sample Recall@50 | Mixed |
| lower acceptance threshold (ablation) | SpecGR++ · Amazon Games | In-sample Recall@50 | Hurt |
| plug SpecGR into DSI backbone with different drafters | DSI · Amazon Games | NDCG@50 overall: 0.0198 -> 0.0220 | Improved |
Closest papers by meaning
- Actions Speak Louder than Words: Trillion-Parameter Sequential Transducers for Generative Recommendations (ICML 2024 · similarity 0.90)
- Rethinking Semantic ID Construction for Generative Recommendation: SimHash with Parallel Decoding and Semantic Alignment (NeurIPS 2026 · similarity 0.88)
- Echoes in Filter Bubble: Diagnosing and Curing Popularity Bias in Generative Recommender Systems (NeurIPS 2026 · similarity 0.88)
- Efficient Inference for Large Language Model-based Generative Recommendation (ICLR 2025 · similarity 0.87)
- Sparse Meets Dense: Unified Generative Recommendations with Cascaded Sparse-Dense Representations (NeurIPS 2025 · similarity 0.86)
- Preference Diffusion for Recommendation (ICLR 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
