Research map / Large language models / AI and society
Telepath: Understanding Users from a Human Vision Perspective in Large-Scale Recommender Systems
Yu Wang, Jixing Xu, Aohan Wu, Mantian Li, Yang He, Jinghe Hu, Weipeng Yan
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 image-based vision extraction and interest understanding rather than item-ID embeddings | Telepath · Offline, image-only experiment | AUC: 0.664 -> 0.693 | Improved |
| Use image-based vision extraction and interest understanding rather than item-ID embeddings | Telepath · Offline, image-only experiment | log loss: 0.675 -> 0.665 | Improved |
| Add Telepath's vision and interest modules to full-featured Wide & Deep | Wide & Deep · Offline, full-featured experiment | AUC: 0.701 -> 0.718 | Improved |
| Add Telepath's vision and interest modules to full-featured Wide & Deep | Wide & Deep · Offline, full-featured experiment | log loss: 0.642 -> 0.627 | Improved |
| Use Telepath ranking instead of the previous ranking model | JD app item recommendation block X · JD app item recommendation block X | CTR: +1.59% | Improved |
| Use Telepath ranking instead of the previous ranking model | JD app item recommendation block X · JD app item recommendation block X | GMV: +8.16% | Improved |
| Use Telepath ranking instead of the previous ranking model | JD app item recommendation block X · JD app item recommendation block X | orders: +8.71% | Improved |
| Use Telepath instead of the previous ranking model | JD DSP publishers Y, first launch · JD DSP publishers Y | GMV: +61.72% average; -19.24% on Day4 | Mixed |
| Use Telepath instead of the previous ranking model | JD DSP publishers Y, first launch · JD DSP publishers Y | ROI: +65.57% average; -17.44% on Day4 | Mixed |
| Use all three ad images instead of only the first image for three-image ads (ablation) | Telepath, version 1 · JD DSP publishers Y, second launch | GMV: +41.75% | Improved |
| Use all three ad images instead of only the first image for three-image ads (ablation) | Telepath, version 1 · JD DSP publishers Y, second launch | ROI: +41.37% | Improved |
| Feed the browse sequence in inverse order to the RNN rather than forward order (ablation) | Telepath interest understanding module · Offline, image-only experiment | AUC | Improved |
Closest papers by meaning
- Recommender Ecosystems: A Mechanism Design Perspective on Holistic Modeling and Optimization (AAAI 2024 · similarity 0.83)
- Think Wise, Collaborate Effectively: A Rationale-Aware LLM-Based Recommender with Reinforcement Learning from Collaborative Signals (AAAI 2026 · similarity 0.83)
- Knowledge-Enhanced Hierarchical Graph Transformer Network for Multi-Behavior Recommendation (AAAI 2021 · similarity 0.82)
- Actions Speak Louder than Words: Trillion-Parameter Sequential Transducers for Generative Recommendations (ICML 2024 · similarity 0.82)
- UxSID: Semantic-Aware User Interests Modeling for Ultra-Long Sequence (NeurIPS 2026 · similarity 0.82)
- LLM-Powered User Simulator for Recommender System (AAAI 2025 · similarity 0.82)
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.
