Auto Added by WPeMatico
How do you turn slow, manual click work across browsers and desktops into a reliable, automated system that can actually use a computer for you at scale? Lux is the latest example of computer use agents moving from research demo to infrastructure. OpenAGI Foundation team has released Lux, a foundation model that operates real desktops […]
How do you keep RAG systems accurate and efficient when every query tries to stuff thousands of tokens into the context window and the retriever and generator are still optimized as 2 separate, disconnected systems? A team of researchers from Apple and University of Edinburgh released CLaRa, Continuous Latent Reasoning, (CLaRa-7B-Base, CLaRa-7B-Instruct and CLaRa-7B-E2E) a
NVIDIA announced today a significant expansion of its strategic collaboration with Mistral AI. This partnership coincides with the release of the new Mistral 3 frontier open model family, marking a pivotal moment where hardware acceleration and open-source model architecture have converged to redefine performance benchmarks. This collaboration is a massive leap in inference speed: the
Large language model agents are starting to store everything they see, but can they actually improve their policies at test time from those experiences rather than just replaying context windows? Researchers from University of Illinois Urbana Champaign and Google DeepMind propose Evo-Memory, a streaming benchmark and agent framework that targets this exact gap. Evo-Memory evaluates
How do you get GPT-5-level reasoning on real long-context, tool-using workloads without paying the quadratic attention and GPU cost that usually makes those systems impractical? DeepSeek research introduces DeepSeek-V3.2 and DeepSeek-V3.2-Speciale. They are reasoning-first models built for agents and targets high quality reasoning, long context and agent workflows, with open weights and production APIs. The
How do you keep synthetic data fresh and diverse for modern AI models without turning a single orchestration pipeline into the bottleneck? Meta AI researchers introduce Matrix, a decentralized framework where both control and data flow are serialized into messages that move through distributed queues. As LLM training increasingly relies on synthetic conversations, tool traces
Why do current audio AI models often perform worse when they generate longer reasoning instead of grounding their decisions in the actual sound. StepFun research team releases Step-Audio-R1, a new audio LLM designed for test time compute scaling, address this failure mode by showing that the accuracy drop with chain of thought is not an
How can an AI system learn to pick the right model or tool for each step of a task instead of always relying on one large model for everything? NVIDIA researchers release ToolOrchestra, a novel method for training a small language model to act as the orchestrator- the ‘brain’ of a heterogeneous tool-use agent https://arxiv.org/pdf/2511.21689
How can an AI system prove complex olympiad level math problems in clear natural language while also checking that its own reasoning is actually correct? DeepSeek AI has released DeepSeekMath-V2, an open weights large language model that is optimized for natural language theorem proving with self verification. The model is built on DeepSeek-V3.2-Exp-Base, runs as
AI applications rarely deal with one clean table. They mix user profiles, chat logs, JSON metadata, embeddings, and sometimes spatial data. Most teams answer this with a patchwork of an OLTP database, a vector store, and a search engine. OceanBase released seekdb, an open source AI focused database (under the Apache 2.0 license). seekdb is
