Research map / Large language models / Language, safety and interpretability
CRAMER: Control via Request-Aware Masking for Editing Recommenders
Zhiyuan Su, Naihe Feng, Luther Qin, Ga Wu
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 CRAMER request-aware masking to frozen SASRec | SASRec · KuaiSAR | HR@20: 0.601 -> 0.748 | Improved |
| add CRAMER request-aware masking to frozen BERT4Rec | BERT4Rec · CDs&Vinyl | NDCG@10: 0.300 -> 0.487 | Improved |
| replace REARANK LLM reranking with CRAMER masking | SASRec · ReDial | HR@10: 0.549 -> 0.578 | Improved |
| replace REARANK with CRAMER on BERT4Rec | BERT4Rec · Beauty | MRR@10: 0.331 -> 0.345 | Improved |
| replace REARANK with CRAMER on SASRec | SASRec · Beauty | MRR@10: 0.335 -> 0.385 | Improved |
| replace LLM-ESR with CRAMER on BERT4Rec | BERT4Rec · CDs&Vinyl | HR@20: 0.692 -> 0.695 | Improved |
| use CRAMER instead of REARANK for inference cost | SASRec · Inference efficiency | added runtime (s): +9.256 -> +0.018 | Improved |
| increase BERT4Rec backbone depth from 4 to 10 layers with CRAMER (ablation) | CRAMER on BERT4Rec · KuaiSAR | NDCG@10: 0.434 -> 0.462 | Improved |
| issue request 'I'd like a romantic comedy' | CRAMER · ReDial romance share in top-10 | romance proportion: 0.286 -> 0.432 | Improved |
| issue request 'Please avoid romantic movies' | CRAMER · ReDial romance share in top-10 | romance proportion: 0.286 -> 0.135 | Improved |
| sample gate masks independently per instance instead of sharing across batch (ablation) | CRAMER · all datasets, both backbones | NDCG@10 | Improved |
| use cosine instead of linear/exponential temperature annealing (ablation) | CRAMER · all datasets | NDCG@10 | Improved |
Closest papers by meaning
- LLM4RSR: Large Language Models as Data Correctors for Robust Sequential Recommendation (AAAI 2025 · similarity 0.88)
- SLMRec: Distilling Large Language Models into Small for Sequential Recommendation (ICLR 2025 · similarity 0.87)
- MARS: Multi-resolution Adaptive Routing for Sequential Recommendation (NeurIPS 2026 · similarity 0.86)
- LLM-ESR: Large Language Models Enhancement for Long-tailed Sequential Recommendation (NeurIPS 2024 · similarity 0.86)
- Actions Speak Louder than Words: Trillion-Parameter Sequential Transducers for Generative Recommendations (ICML 2024 · similarity 0.86)
- Objective Shaping with Hard Negatives: Windowed Partial AUC Optimization for RL-based LLM Recommenders (NeurIPS 2026 · 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.
