Retrieval
PulseAugur coverage of Retrieval — every cluster mentioning Retrieval across labs, papers, and developer communities, ranked by signal.
5 day(s) with sentiment data
-
AI agent costs surge due to lack of granular expense tracking
As AI agents become more complex, their associated costs are increasing significantly, with one team experiencing a fourfold rise in inference spend. The core issue is a lack of granular cost attribution, as current mon…
-
RAG vs. Fine-Tuning: Choosing the Right LLM Strategy Based on Failure Type
This article explores the strategic choices between retrieval-augmented generation (RAG) and fine-tuning for large language models, suggesting a layered approach based on the type of failure encountered. Prompting is pr…
-
AI stack complexity extends beyond models to data, inference, and observability
The complexity of AI systems extends beyond the core model itself, encompassing a broader stack of components. Key elements such as data management, retrieval mechanisms, context handling, inference processes, evaluatio…
-
New CAPE method restores long-tail item recall in generative recommenders
A new arXiv paper introduces CAPE (Co-occurrence-Aware Pool Expansion), a method designed to improve generative recommender systems by restoring their ability to recall long-tail items. Current generative models often f…
-
Enhancing RAG Systems with User Intent and Smarter Retrieval Techniques
This article delves into enhancing Retrieval-Augmented Generation (RAG) systems by focusing on user intent. It explores techniques such as intent detection, self-querying retrieval, metadata filtering, and query rewriti…
-
RAG vs. Fine-Tuning: Choosing the Right AI Chatbot Enhancement
Retrieval-Augmented Generation (RAG) is a technique that enhances chatbots by allowing them to access and cite external documents at the time of a user's query, rather than requiring extensive retraining. This method is…
-
Chunking Dilemma: Optimizing Document Splits for RAG Systems
The process of preparing documents for retrieval-augmented generation (RAG) involves breaking down large texts into smaller, manageable pieces called chunks. This 'chunking' is crucial because simply feeding an entire d…
-
Advanced RAG techniques enhance retrieval accuracy and vector database integration
This series of articles delves into advanced techniques for Retrieval-Augmented Generation (RAG), focusing on improving retrieval accuracy and efficiency. The first article explores semantic chunking, parent-document re…
-
New AutoDataBench testbed isolates AI agent data intelligence
Researchers have introduced AutoDataBench, a new testbed designed to isolate and evaluate "Data Intelligence" in AI agents. This framework aims to disentangle the impact of data improvements from other factors like trai…
-
AI text detection faces fundamental collision-entropy floor, researcher claims
A researcher has proposed a formalization suggesting that similarity-based AI-text detectors, such as watermarking and retrieval methods, face a fundamental limitation in their false-positive rate. The core argument pos…
-
LLMs vs. Embedding Models: Costly Parity Found in New Study
A new paper titled "The Embedder's Dilemma" compares the performance and cost of large language models (LLMs) against dedicated embedding models for various tasks. The study found that while LLMs like Gemini 3.1 Pro per…
-
Feature stores solve ML training-serving drift and target leakage
Feature stores are essential for production machine learning, addressing the critical problem of feature definition drift between training and serving environments. This drift can silently degrade model performance, par…
-
DART architecture enhances long-context sequence modeling by merging Transformers and SSMs
Researchers have introduced DART (Decoded Attention over Recurrent States), a novel architecture that combines the strengths of Transformers and State Space Models (SSMs) for efficient long-context sequence modeling. DA…
-
Cologne Information Retrieval group presents agentic conversational search system
Researchers from the Cologne Information Retrieval group have developed an agentic conversational search system for the iKAT SCAI 2026 shared task. This system incorporates tools for query rewriting, retrieval, rerankin…
-
New research questions spectral analysis for time-series forecasting
Two recent arXiv papers explore the limitations of using spectral analysis for time-series forecasting, particularly when incorporating external context. The first paper introduces CAF-7M, a large dataset designed to im…
-
New benchmarks evaluate Portuguese text embedding models, revealing performance gaps
Two new benchmarks, MTEB-PT and MTEB-PT (Brazilian Portuguese), have been released to evaluate text embedding models specifically for the Portuguese language. These benchmarks address the underrepresentation of Portugue…
-
LLM context windows are not knowledge bases; strategic selection is key
The article argues that simply increasing the context window size of an LLM does not equate to effective context engineering. Instead, it emphasizes the importance of strategically selecting and presenting the most rele…