Technical AI news, automatically curated and generated
Manifest has officially deprecated its LLM router, a feature originally launched in March as part of its LLM gateway to reduce inference costs. The router functioned by classifying incoming requests into four complexity …
→ View original source
Researchers have introduced Σ-Mem, an online reliability memory framework designed to enhance coordination in LLM-based multi-agent systems by modeling agent competence rather than just interaction history. Unlike tradit…
→ View original source
This research investigates the mechanisms driving sparse Mixture-of-Experts (MoE) routing, challenging the assumption that expert benefits stem solely from geometric complementarity. By introducing the Expert Subspace Se…
→ View original source
This research investigates the implications of lossy verification in Speculative Decoding (SD), a technique designed to accelerate large language model inference by utilizing a lightweight draft model to propose tokens f…
→ View original source
This research presents the first systematic investigation into filesystem-based memory for Large Language Model (LLM) agents, treating a directory tree of Markdown files as a unified storage medium for declarative memory…
→ View original source
DeepSeek has officially released the DeepSeek-V4.1-Flash model, the smallest entry in a new architecture family designed for higher throughput, faster inference, and expanded scaling potential. This new architecture feat…
→ View original sourceDeepSeek V4 Flash 0731 (Reasoning, Max Effort) is a large-scale Mixture of Experts (MoE) model released on July 31, 2026, featuring 284 billion total parameters with only 13 billion active parameters during inference. De…
→ View original source
This technical overview details the engineering requirements for implementing a browser-local video face-swapping pipeline using WebGPU, WebCodecs, and ONNX Runtime Web. To ensure usability, the architecture avoids proce…
→ View original source
The Model Context Protocol (MCP) is an open standard designed to solve the fragmentation caused by developers needing to build individual integrations for every AI assistant and service. Instead of creating bespoke conne…
→ View original sourceOpenAI has announced significant price and performance updates to the GPT-5.6 model family, driven by optimizations across model architecture, inference systems, and agentic harnesses. To improve cost-efficiency for ente…
→ View original source
The Qwen-UI-Agent technical report introduces a foundation GUI agent designed for real-world digital environments, including mobile, computer-use, web, and DeepSearch platforms. To bridge the gap between simulation and r…
→ View original source
Researchers have introduced Fairness Pruning, a lightweight structural intervention designed to localize demographic bias within the Gated Linear Unit (GLU) MLP layers of large language models. By employing minimally con…
→ View original source