Research map / Vision-language and multimodal / Human and embodied modeling
Learning Instance-Level Representation for Large-Scale Multi-Modal Pretraining in E-Commerce
Yang Jin, Yongzhi Li, Zehuan Yuan, Yadong Mu
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 ECLIP instance-centric pretraining instead of global image-text pretraining | ECLIP · M5Product zero-shot classification | Acc@1: 39.3 -> 44.8 | Improved |
| Use ECLIP instance-centric pretraining instead of global image-text pretraining | ECLIP · Fine-grained product retrieval | R@1: 35.6 -> 45.0 | Improved |
| Add inter-product multi-modal learning (ablation) | ECLIP · Coarse-level product retrieval | mAP@1 (%): 61.6 -> 64.4 | Improved |
| Add intra-product multi-modal learning (ablation) | ECLIP · Coarse-level product retrieval | mAP@1 (%): 61.6 -> 62.2 | Improved |
| Combine inter-product and intra-product learning (ablation) | ECLIP · Coarse-level product retrieval | mAP@1 (%): 64.4 -> 65.2 | Improved |
| Use random negative queries instead of negative text prompts or EMA-updated queries (ablation) | ECLIP · Instance-level product retrieval | mAP@10: 84.9; 85.1; 86.8 | Mixed |
| Increase instance query count from 20 to 40 (ablation) | ECLIP · Instance-level product retrieval (visual modality) | mAP@10: 89.6 -> 89.9 | Improved |
| Replace cross-attention with slot-attention (ablation) | ECLIP decoder · Visual grounding | Acc@0.7: 66.3 -> 70.5 | Improved |
| Add instance-text matching loss (ablation) | ECLIP · Visual grounding | Acc@0.7: 69.6 -> 70.0 | Improved |
| Add entropy regularization (ablation) | ECLIP with instance-text matching loss · Visual grounding | Acc@0.7: 70.0 -> 70.5 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- BLIP — BLIP: Bootstrapping Language-Image Pre-training for Unified Vision-Language Understanding and Generation (ICML 2022) · better on 6 of 6 shared results
- FILIP — FILIP: Fine-grained Interactive Language-Image Pre-Training (ICLR 2022) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- eCeLLM: Generalizing Large Language Models for E-commerce from Large-scale, High-quality Instruction Data (ICML 2024 · similarity 0.85)
- EcomGPT: Instruction-Tuning Large Language Models with Chain-of-Task Tasks for E-commerce (AAAI 2024 · similarity 0.84)
- TIGER-FG: Text-Guided Implicit Fine-Grained Grounding for E-commerce Retrieval (NeurIPS 2026 · similarity 0.84)
- Transformer-Empowered Multi-Modal Item Embedding for Enhanced Image Search in E-commerce (AAAI 2024 · similarity 0.84)
- Towards Unified Multi-Modal Personalization: Large Vision-Language Models for Generative Recommendation and Beyond (ICLR 2024 · similarity 0.83)
- Cross-modal Representation Flattening for Multi-modal Domain Generalization (NeurIPS 2024 · 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.
