MahaBodi, a Rust-based enterprise router from Laya, decides among up to 100,000 candidates in a single forward pass and beat Laya with a dense shortlist on entity linking at 100K Wikipedia pages (0.17…
DSPy 3.4 adds first-class support for System One models, letting Jev run behind an ordinary DSPy signature and preserving its probability evidence for optimization. The integration installs via `pip i…
A developer compared five approaches to a banking-support classification task using the BANKING77 dataset: a TF-IDF lexical model, a small sentence transformer (MiniLM), a general-purpose language mod…
Researchers Kostia Kudriavtsev, Parvez Rafi, and Sha Sundaram presented Glyph, a production multi-agent LLM system that generates column descriptions and assigns sensitivity-ontology tags for enterpri…
A developer has released EmbedFlow, an open-source tool that enables zero-downtime upgrades of embedding models by retrieving top-K documents with the old model and reranking them with the new model, …
A developer building a RAG agent for a scanned Kannada novel found that multilingual embeddings underperform on agglutinative Kannada, leading to retrieval failures. By implementing a hybrid BM25 + de…
A developer detailed a sequential model-loading approach to run Whisper, bge-m3, and Gemma on a single local workstation with limited VRAM, avoiding out-of-memory errors by loading and unloading model…
OpenAI's GPT-5.4-mini can classify e-commerce queries more cheaply by hallucinating fake categories and matching them to real ones via embeddings, according to a developer who shared the technique. Th…
Developer Siddhartha Katiyar built ContextOS, a local-first context engine that indexes software repositories using Tree-sitter and SQLite FTS5 to improve AI coding assistants' retrieval accuracy. In …
A developer built a 3-tier on-device AI concierge widget that runs entirely in the visitor's browser, using Google's Gemini Nano, a MiniLM embedding model, and keyword matching, achieving zero cost pe…
An engineer built NilaMind, an open-source Android app that runs a 1.5B-parameter language model entirely on-device for mental health conversations, with no cloud, account, or analytics. The app uses …
A developer built a browser instrument that visualizes how embedding coordinates can change under orthogonal transformations while cosine similarities remain identical. The tool demonstrates that raw …
A new retrieval-augmented generation system, LongEval-RAG, outperforms complex alternatives by using a simple methodology, challenging the AI community's focus on intricate designs. The rule-minilm va…
Researchers developed a new method for out-of-scope intent detection using MiniLM embeddings, treating it as a one-class classification task with multi-cluster boundary learning. The approach outperfo…
Researchers propose a multi-cluster boundary learning method for out-of-scope intent detection using MiniLM embedding, achieving state-of-the-art performance on CLINC150, StackOverflow, and Banking77 …
Researchers solved BlueDot's TAIS Puzzle #1 by identifying a nonlinear representation of the 'country' feature in a five-layer MLP, hidden as an XOR with 'food' at layer h2. They used Distributed Alig…
Rememori, a zero-dependency agent memory engine written in pure TypeScript, launched on Hacker News. The library provides memory primitives for AI agents via a three-verb API (remember, recall, forget…
Independent researcher developed ZATRON, a system that encrypts semantic search embeddings while preserving 98% search quality, preventing attackers from clustering documents by topic. Tested on MSMAR…
Rampart, a 14.7MB on-device personal information filtering system, was open-sourced to redact PII before it leaves a user's device. It combines deterministic rules and a MiniLM model to catch structur…
MemDelta, a controlled evaluation protocol for agent memory systems, reveals that reported performance gains often stem from changes in embedding models or language models rather than memory architect…