LLM-Based Query Expansion with Gaussian Kernel Semantic Enhancement for Dense Retrieval (Electronics, MDPI, 2025)

Forum topic · 小凯 · 2026-07-05

Summary

This forum post catalogs an academic paper published in MDPI Electronics (May 2025 issue, article 1744): "LLM-Based Query Expansion with Gaussian Kernel Semantic Enhancement for Dense Retrieval." The paper addresses query expansion for dense retrieval, a long-standing challenge in information retrieval, by using large language models to expand queries and applying a Gaussian kernel–based semantic enhancement to refine the expanded representations. The post situates the work within the broader Query Understanding section of an awesome list covering LLM-era search, recommendation, and personalization systems, noting the shift from pipeline-based retrieval toward architectures that integrate LLM reasoning, retrieval, and generation. It also cross-references related work on query expansion, including Aligned Query Expansion, LLM-trained query rewriting, tip-of-the-tongue query decomposition, and few-shot conversational query rewriting. The post includes engineering checklists for deploying retrieval systems (latency budgets, data privacy, cost control, safety) and a glossary of IR/RAG terminology. Readers should consult the original PDF for exact experimental numbers.

LLM-Based Query Expansion with Gaussian Kernel Semantic Enhancement for Dense Retrieval (MDPI Electronics, 2025)

This post introduces a peer-reviewed paper from the Query Understanding section of the forum's information-retrieval collection.

Metadata

| Field | Value | |-------|-------| | Title | LLM-Based Query Expansion with Gaussian Kernel Semantic Enhancement for Dense Retrieval | | Venue | MDPI *Electronics*, Vol. 14, Issue 9, Article 1744 (March 2025 submission) | | Source | https://www.mdpi.com/2079-9292/14/9/1744 | | Resource type | Academic paper | | Section | Query Understanding |

One-line summary

The paper combines LLM-based query expansion with a Gaussian kernel semantic enhancement mechanism to improve dense retrieval quality.

Context and motivation

Large-scale search, recommendation, and personalization systems have long faced challenges in efficiency, scalability, and user-intent understanding. Traditional pipelines separate retrieval, ranking, and generation, which struggles to meet the demands of the LLM era: natural-language interaction, multi-hop reasoning, and up-to-date knowledge. This paper targets one core sub-problem — query expansion for dense retrieval — where the quality of the query representation directly determines recall and ranking effectiveness.

Method (as described by the post)

Following the typical four-step pattern of problem formalization → model/system design → training/construction → inference pipeline:

1. Input & representation: encode the query, documents, and user context into dense representations or structured prompts; 2. Core module: LLM-based query expansion, enhanced by a Gaussian kernel weighting of semantics to control how expansion terms contribute; 3. Learning strategy: standard dense-retrieval training with contrastive objectives; 4. Inference: single-pass or iterative retrieval with dense nearest-neighbor search.

*Note: exact technical details and quantitative results should be verified against the original PDF; this post is based on the abstract and public metadata.*

Typical evaluation setup for this problem area

Datasets: MS MARCO, BEIR, Natural Questions, domain-specific corpora
Metrics: nDCG@10, MRR, Recall@k
Baselines: BM25, vanilla dense retrievers, cross-encoder rerankers, retrieval-free LLMs

Key takeaways for Search / Rec / Personalization

1. Architecture: cascaded retrieve + rerank + generate remains mainstream, but agentic paradigms are making retrieval strategy itself learnable; 2. Data: high-quality instruction data and click/session logs matter as much as model design; synthetic data requires guarding against knowledge leakage; 3. Evaluation: the gap between offline metrics and online satisfaction is widening; LLM-as-judge needs cross-validation with human assessment; 4. Productization: latency, cost, explainability, and safety are hard constraints for industrial deployment.

Engineering checklist

| Item | Question | Suggestion | |------|----------|------------| | Data | PII in training/index? Version control? | Partitioned indexes, anonymization, rollback-capable embedding versions | | Latency | p99 budget? Retrieval steps? | Cascade + early stopping, hot-query caching, async reranking | | Quality | Does offline gain translate to online CTR? | Interleaving experiments, human audits, citation verification | | Safety | Poisoning/bias from open retrieval? | Source whitelists, adversarial detection, output filtering | | Cost | Token and GPU usage per query? | Small-model routing, distillation, hybrid sparse+dense |

Related entries

Aligned Query Expansion: Efficient Query Expansion for Information Retrieval (arXiv:2507.11042)
Beyond the Limitation of a Single Query: Train Your LLM for Query Expansion (arXiv:2510.10009)
Decomposing Complex Queries for Tip-of-the-tongue Retrieval (arXiv:2305.15053)
Few-Shot Generative Conversational Query Rewriting (SIGIR 2020, DOI: 10.1145/3397271.3401323)
Hierarchical Query Classification in E-commerce Search (WWW 2024)
Hypothetical Documents or Knowledge Leakage? Rethinking LLM-based Query Expansion (arXiv:2504.14175)

Glossary

| Term | Meaning | |------|---------| | IR | Information Retrieval | | RAG | Retrieval-Augmented Generation | | LTR | Learning to Rank | | nDCG | Normalized Discounted Cumulative Gain | | Agentic Search | Modeling search as sequential decision-making and tool use | | Gen-IR | Generative Information Retrieval |