Enhancing Relevance of Embedding-based Retrieval at Walmart (CIKM 2024)
This post is a forum index entry for the academic paper "Enhancing Relevance of Embedding-based Retrieval at Walmart", presented at CIKM 2024 (October 2024).
Metadata
| Field | Content | |-------|---------| | Title | Enhancing Relevance of Embedding-based Retrieval at Walmart | | Venue | CIKM 2024 (October 2024) | | Type | Academic paper | | Source | https://dl.acm.org/doi/abs/10.1145/3627673.3680047 |
Context
The paper addresses relevance in large-scale embedding-based retrieval within Walmart's search/recommendation stack. In modern search systems, dense embeddings power candidate recall, but maintaining query–result relevance remains challenging due to domain shift, long-tail queries, and the shift toward LLM-era architectures where retrieval, ranking, and generation must be coordinated.
The forum commentary situates the work along the standard pipeline:
1. Input & representation — encoding queries, documents, and user context into dense/sparse representations; 2. Core modules — retrievers, re-rankers, planners, and feedback mechanisms; 3. Learning strategies — supervised fine-tuning, contrastive learning, distillation, and reinforcement learning; 4. Inference strategies — single-pass retrieval, iterative retrieval, parallel sub-queries, and early stopping with budget control.
Engineering checklist highlights
| Concern | Recommendation | |---------|----------------| | Latency | Cascaded retrieval + early stopping, query caching, async re-ranking | | Quality | Interleaving experiments, human audits, citation verification | | Safety | Source whitelisting, adversarial detection, output filtering | | Cost | Model routing, distillation, hybrid sparse+dense retrieval |
Related entries
- An interpretable ensemble of graph and language models
- Applying Deep Learning to Ads Conversion Prediction in Last Mile Delivery (arXiv:2502.10514)
- Automated Query-Product Relevance Labeling using Large Language Models (arXiv:2502.15990)
- Behavior Modeling Space Reconstruction for E-Commerce Search (arXiv:2501.18216)
- Behavior-driven query similarity prediction based on pre-trained language models
- Better to Ask in English: Cross-Lingual Evaluation of Large Language Models (DOI: 10.1145/3589334.3645643)
Notes
The original post does not reproduce the paper's abstract text or experimental tables; it primarily provides metadata, context, and a glossary (IR, RAG, LTR, nDCG, agentic search, generative IR). For quantitative results, refer to the original paper on ACM DL.