Tagged articles
semantic retrieval
23 articles · Page 1 of 1

XikeSep 14, 2026 · Artificial Intelligence
This article explains intent recognition for AI agents, covering definition, common approaches (rules, LLM with structured output, semantic retrieval, hybrid), their pros and cons, suitable scenarios, and practical implementation advice including schema validation, confidence gating, and multi-turn handling.
AI agentsFunction CallingLLM
0 likes · 15 min read

The article explains that effective conversation memory for LLMs requires classifying information into static knowledge, short‑term context, and long‑term memory, defining a full lifecycle for each entry, and implementing strict storage, retrieval, update, and deletion policies rather than simply concatenating all history or keeping only the latest turns.
Conversation MemoryLLMMemory Lifecycle
0 likes · 24 min read

Mandol tackles the fragmented memory problem of LLM agents by unifying representation, storage, and retrieval in a memory‑native architecture; benchmarked on LoCoMo and LongMemEval it achieves up to 92.21% accuracy, 5× faster latency, and runs efficiently on consumer‑grade hardware without external databases.
Agent MemoryLLMhierarchical memory
0 likes · 14 min read

The article surveys memory mechanisms across major Agent Harness frameworks, classifies three memory types, evaluates each system’s implementation, highlights benchmark shortcomings, and presents Mem0 as a unified solution that overcomes capacity, retrieval, and isolation limitations.
AI agentsAgent HarnessMem0
0 likes · 19 min read

The article examines how Alibaba Cloud OSS Vector Bucket solves the data‑scattered, costly, and inefficient retrieval problems of massive multimodal media asset platforms by unifying storage, providing semantic vector search, and cutting operational expenses up to 95%.
Multimodal DataOSS Vector BucketVector Search
0 likes · 9 min read

The article explains how AI agents use memory to retain conversation context, why sending the full history to large language models is inefficient, and presents rolling context windows, inverted‑index pruning, semantic embedding retrieval, and GraphRAG as complementary strategies to build more accurate and personalized agents.
AI MemoryGraphRAGLLM Optimization
0 likes · 10 min read

The article explains how Hypothetical Document Embeddings (HyDE) improve Retrieval‑Augmented Generation by generating a synthetic answer before vector search, allowing the system to embed richer semantic intent rather than relying on shallow keyword similarity, and provides a step‑by‑step implementation using LangChain.
HyDELLMLangChain
0 likes · 6 min read

The article explains how Ontology—an abstract model of objects, relationships, and actions—can be built on PolarDB‑PG’s intelligent engine to overcome semantic ambiguity and logical hallucination in enterprise LLM agents, describing a three‑layer architecture, OAG retrieval, automatic modeling, fine‑grained permission control, and real‑world supply‑chain use cases.
AI agentLLMPolarDB-PG
0 likes · 13 min read

The article analyzes why conventional AI‑agent memory—mere context concatenation—fails in long or cross‑session interactions, then proposes a four‑capability, three‑layer hierarchical memory framework (L1‑L3) with concrete design strategies, progressive summarisation, semantic retrieval, conflict resolution and decay mechanisms.
AI agentLLMMemory Architecture
0 likes · 15 min read

RAG (Retrieval‑Augmented Generation) is a hot interview topic, and many candidates stumble on semantic‑loss issues; this article dissects a real JD interview case, identifies three core shortcomings, and presents a three‑step technical solution—structure restoration, semantic splitting, and hybrid retrieval—plus a ready‑to‑use answer template.
AI InterviewRAGdocument parsing
0 likes · 25 min read

This article presents an ontology‑aware knowledge‑graph RAG framework that transforms complex, hierarchical industrial standard documents into a graph of sections, atomic propositions, and refined triples, achieving nearly double F1 scores on table‑based QA tasks and robust performance on long documents.
LLMRAGTable QA
0 likes · 6 min read

This article details how an automotive AI assistant migrated from keyword matching to a Milvus‑based vector retrieval system, overcoming semantic gaps, scaling to millions of daily queries, optimizing indexing, introducing multi‑vector and sparse‑vector search, and building a real‑time RAG pipeline with Flink.
AI assistantMilvusRAG
0 likes · 12 min read

The article explains how the limited context window of large language models causes prompt bloat when many tool descriptions are embedded, and presents the RAG‑MCP architecture that stores tool metadata in a vector database, uses semantic retrieval to select only the most relevant tools, dramatically shortens prompts, and improves inference speed and tool‑call accuracy.
Amazon BedrockLLMMCP
0 likes · 25 min read

The article examines eight common AI memory techniques—from simple full‑history retention to sophisticated vector‑store and knowledge‑graph approaches—detailing their principles, Python‑style implementations, advantages, drawbacks, and ideal application scenarios for large‑language‑model agents in production environments.
AI MemoryLLM context managementknowledge graph
0 likes · 23 min read

Context Engineering, a new discipline for optimizing large language model inputs, expands context windows, compares with prompt engineering, outlines core techniques like information organization, dynamic management, semantic retrieval, and offers practical applications and recommendations to enhance AI performance across domains.
ai-optimizationlarge language modelsprompt engineering
0 likes · 11 min read

This article presents a method for massive multi-label text classification on Zhihu content by combining a semantic retrieval model with a proprietary large AI model, detailing the challenges of large label spaces, model architecture, loss optimization, and experimental results showing significant accuracy gains.
BGEZhihularge language model
0 likes · 16 min read

Xiaohongshu’s search advertising recall system evolves from keyword bidding to BERT‑based vector retrieval and LLM‑enhanced query rewriting, using dual semantic and efficiency models, water‑level metrics, and GPU‑accelerated engineering to achieve 80 % click coverage, 60 % conversion coverage and a 5 % CPM lift.
artificial intelligenceefficiency optimizationlarge language models
0 likes · 33 min read

The article chronicles Baidu Search’s Model Architecture Group’s evolution of deep‑learning‑driven search, detailing the shift from inverted‑index to semantic vector indexing, the use of transformer‑based models for text and image queries, large‑scale offline/online pipelines, and extensive GPU‑centric optimizations such as pruning, quantization and distillation, all aimed at delivering precise, cost‑effective results to hundreds of millions of users.
ERNIEGPU InferenceSearch Architecture
0 likes · 14 min read

Baidu Netdisk’s new image search combines ERNIE‑ViL‑based semantic vectors, cross‑modal matching and metadata such as timestamps, GPS and facial tags, using LSH‑optimized indexing to let users find specific photos among billions with natural‑language queries, delivering faster, more accurate results without manual tagging.
ERNIE-ViLLSH hashingimage search
0 likes · 11 min read

The Baidu Search Closed‑Door Technical Symposium, the first core technical forum hosted by Baidu’s Search Architecture Department, brings senior engineers and junior backend developers together to discuss semantic retrieval, data‑driven big‑data processing, and vertical search offline architecture, while offering limited‑capacity sessions, networking gifts, and travel subsidies.
Baidu SearchCloud-Native Architecturebackend development
0 likes · 6 min read

This article describes how Dingxiangyuan's algorithm team adopted Milvus for distributed vector indexing to improve semantic search in their community forum, detailing the background, retrieval workflow, various embedding models—including Bi‑Encoder, Spherical Embedding, and Knowledge Embedding—and summarizing the benefits and future applications.
EmbeddingMilvusNLP
0 likes · 10 min read

This talk by Zhihu search algorithm engineer Shen Zhan details the evolution of text relevance models from TF‑IDF/BM25 to deep semantic matching and BERT, explains the challenges of deploying BERT at scale, and describes practical knowledge‑distillation techniques that improve both online latency and offline storage while maintaining search quality.
BERTknowledge distillationmachine learning
0 likes · 14 min read

This article details Zhihu's search engineering advances over the past year, covering long‑tail query challenges, term‑weight calculation, synonym expansion, query rewriting with translation models and reinforcement learning, and semantic retrieval using BERT‑based embeddings, while outlining future research directions.
NLPQuery Rewritingquery understanding
0 likes · 14 min read
