AI Briefing#retrieval
The latest AI and developer news about #retrieval, with the original source and a short summary.
The latest AI and developer news about #retrieval, with the original source and a short summary.
The open-source memory layer memv has launched an MCP server, making it available for use in Claude Desktop, Code, and more.
A new memory retrieval system achieved SOTA on the LongMemEval benchmark using Gemini Flash.
IBM Granite released 97M and 311M multilingual embedding R2 models under Apache 2.0.
vLLM Factory, an open-source tool that efficiently serves encoder models such as GLiNER and ColBERT via a vLLM plugin, has been released.
Subquadratic touted a 12 million token context window with SSA and an 83 score on MRCR v2, surpassing GPT-5.5.
A reference implementation of the five-stage structure of context engineering, built with a Bedrock example.
A log-compress-atomize-link structure improved OpenClaw's memory reusability.
Vercel proposed a cheaper, easier-to-debug knowledge agent that uses the filesystem and bash instead of a vector DB.
PlugMem converts agent logs into knowledge units, enabling more accurate recall with fewer tokens.
Noting that the optimal chunk size differs for each query, combining chunks of multiple sizes improved RAG retrieval performance by up to 37%.
Hugging Face has unveiled RTEB, a new retrieval embedding benchmark designed to precisely measure a model's generalization ability.
A detailed guide on training and fine-tuning Sparse Embedding models using the Sentence Transformers library.
The Qwen3-based embedding and reranker series has boosted multilingual retrieval performance.
Cursor is moving toward **next action prediction** that anticipates even the next keystroke and action.
Embedding quantization techniques are presented as a way to reduce the memory usage and cost of vector search while increasing speed.
LangChain analyzed LangSmith data to reveal 2023 trends in LLM application development and testing.