Research map / Label-efficient and robust learning / Self-supervised and contrastive
AdaptSSR: Pre-training User Model with Augmentation-Adaptive Self-Supervised Ranking
Yang Yu, Qi Liu, Kai Zhang, Yuren Zhang, Chao Song, Min Hou, Yuqing Yuan, ZHIhao Ye, ZAIXI ZHANG, Sanshi Lei Yu
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 |
|---|---|---|---|
| use AdaptSSR self-supervised ranking instead of contrastive pre-training (ablation) | Transformer Encoder user model · TTL age prediction (T1) | accuracy: 2.8% | Improved |
| use AdaptSSR self-supervised ranking instead of contrastive pre-training (ablation) | Transformer Encoder user model · TTL thumb-up recommendation (T4) | NDCG@10: 11.8% | Improved |
| add in-batch hard negative sampling (ablation) | AdaptSSR · TTL age prediction (T1) | accuracy: 0.9%-1.6% | Improved |
| add in-batch hard negative sampling (ablation) | AdaptSSR · TTL thumb-up recommendation (T4) | NDCG@10: 3.1%-5.1% | Improved |
| add augmentation-adaptive fusion (ablation) | AdaptSSR · TTL age prediction (T1) | accuracy: 0.9%-1.6% | Improved |
| add augmentation-adaptive fusion (ablation) | AdaptSSR · TTL thumb-up recommendation (T4) | NDCG@10: 3.1%-5.1% | Improved |
| use AdaptSSR rather than the strongest reported baseline | Transformer Encoder user model · TTL age prediction (T1) | accuracy: 2.6% | Improved |
| use AdaptSSR rather than the strongest reported baseline | Transformer Encoder user model · TTL life status prediction (T2) | accuracy: 1.7% | Improved |
| use AdaptSSR rather than the strongest reported baseline | Transformer Encoder user model · TTL click recommendation (T3) | NDCG@10: 6.1% | Improved |
| use AdaptSSR rather than the strongest reported baseline | Transformer Encoder user model · TTL thumb-up recommendation (T4) | NDCG@10: 9.1% | Improved |
| use AdaptSSR rather than the strongest reported baseline | Transformer Encoder user model · App gender prediction (T5) | AUC: 2.3% | Improved |
| use AdaptSSR rather than the strongest reported baseline | Transformer Encoder user model · App CVR prediction (T6) | AUC: 2.6% | Improved |
Closest papers by meaning
- ATRank: An Attention-Based User Behavior Modeling Framework for Recommendation (AAAI 2018 · similarity 0.86)
- HYDRA: Model Factorization Framework for Black-Box LLM Personalization (NeurIPS 2024 · similarity 0.85)
- PANTHER: Generative Pretraining Beyond Language for Sequential User Behavior Modeling (NeurIPS 2025 · similarity 0.85)
- SRA-CL: Semantic Retrieval Augmented Contrastive Learning for Sequential Recommendation (NeurIPS 2025 · similarity 0.84)
- UxSID: Semantic-Aware User Interests Modeling for Ultra-Long Sequence (NeurIPS 2026 · similarity 0.84)
- Augmenting Sequential Recommendation with Balanced Relevance and Diversity (AAAI 2025 · similarity 0.83)
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.
