Research map / Image generation / Autoregressive and latent generation
Probing Product Description Generation via Posterior Distillation
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 APDT instead of KOBE | product description generation · Clothes&Shoes | ROUGE-L: 19.07 -> 20.41 | Helped |
| use APDT instead of KOBE | product description generation · Digital | ROUGE-L: 18.97 -> 19.95 | Helped |
| use APDT instead of KOBE | product description generation · Homing | ROUGE-L: 18.72 -> 19.68 | Helped |
| add a copy mechanism (ablation) | APDT · Clothes&Shoes | ROUGE-L: 19.13 -> 20.41 | Helped |
| add posterior distillation from reviews (ablation) | APDT · Clothes&Shoes | ROUGE-L: 18.56 -> 20.41 | Helped |
| add coherence-enhanced loss (ablation) | APDT · Clothes&Shoes | ROUGE-L: 19.87 -> 20.41 | Helped |
| add the copy mechanism, posterior distillation, and CoE loss together (ablation) | APDT · Clothes&Shoes | ROUGE-L: 17.95 -> 20.41 | Helped |
| distill from the 5-th or 6-th review encoder layer rather than a shallow layer (ablation) | APDT posterior distillation · Clothes&Shoes validation | ROUGE-L | Mixed |
| increase the coherence weight β up to 0.4, then increase it further (ablation) | APDT CoE loss · JDPDG validation | ROUGE-L | Mixed |
Closest papers by meaning
- Aspect-Aware Multimodal Summarization for Chinese E-Commerce Products (AAAI 2020 · similarity 0.84)
- LLM-ESR: Large Language Models Enhancement for Long-tailed Sequential Recommendation (NeurIPS 2024 · similarity 0.81)
- Language Representations Can be What Recommenders Need: Findings and Potentials (ICLR 2025 · similarity 0.81)
- Towards Unified Multi-Modal Personalization: Large Vision-Language Models for Generative Recommendation and Beyond (ICLR 2024 · similarity 0.81)
- SHARP-Distill: A 68× Faster Recommender System with Hypergraph Neural Networks and Language Models (ICML 2025 · similarity 0.81)
- Improving LLMs for Recommendation with Out-Of-Vocabulary Tokens (ICML 2025 · similarity 0.80)
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.