Full-text search across 7151 articles. Combine with topic and date filters; results sorted by relevance.
A new guide outlines four strategies for reducing token usage in multi-agent AI systems: static instruction caching, semantic caching, just-in-time tooling, and task escalation. The guide, published by an unnamed source,…
Practical constraint decoding, also known as structured generation, forces large language models to generate outputs that strictly follow a specified data schema, grammar, or regex by masking logits at the token selectio…
Amazon researchers have developed GraphEval, a framework that uses knowledge graphs to detect hallucinations in large language model (LLM) outputs. The method constructs semantic triples from a model's response and evalu…
Outlines, an open-source library, introduces deterministic certainty into large language model output generation to produce reliable structured outputs such as JSON objects and classification results. The library masks o…
The Center for AI Safety, with world experts, created Humanity's Last Exam (HLE), a benchmark of over 2,500 expert-level questions across disciplines to test AI reasoning. Even top models like GPT, Gemini, and Claude sco…
A developer built a chess bot using a transformer architecture with 11 million parameters, achieving a 1500 Elo rating from the model alone and 2100 Elo when combined with Monte Carlo Tree Search. The bot is available as…
Kimba, a new sleep technology company, launched a clinically validated, AI-powered device that tracks health metrics and releases personalized scents to improve sleep quality. The device, available for preorder at $299, …
PagedAttention, a memory optimization technique in the vLLM inference server, applies virtual memory concepts to manage the KV cache in large language models, improving throughput by reducing fragmentation. The KV cache …
A new study reveals that large language models (LLMs) silently corrupt documents when delegated editing tasks, with even top-tier models like Gemini Pro, Claude Opus, and GPT-5 degrading 25% of original content after 20 …
The field of LLM explainability is advancing with new frameworks like SMILE and gSMILE that use statistical methods to create visual heatmaps showing which parts of a user prompt most influenced a model's output. Researc…
A comprehensive guide has been published covering the history, techniques, datasets, algorithms, tools, and real-world applications of image reconstruction using deep learning. The resource details how deep learning mode…
A mathematician named Norbert Wiener sent a letter in 1949 warning about the dangers of intelligent machines, but his concerns were ignored. The piece traces the history of AI safety concerns from Wiener's cybernetics to…
Given artificial general intelligence (AGI), automating physical production would likely be straightforward because a system capable of all remote cognitive work would also master real-time control, spatial reasoning, an…
Researchers introduced AgentOdyssey, a framework that procedurally generates open-ended text games to evaluate test-time continual learning agents. The framework assesses agents on exploration, knowledge acquisition, mem…
Cheap AI intelligence is repricing labor, shifting work from human execution to machine delegation. Five forecasts predict that prompts become work orders, agents replace chatbots, and companies treat tokens as labor sub…
Researchers Bhavik Chandna and Procheta Sen introduced a counterfactual explanation framework for retrieval models that identifies which terms need to be added to a document to improve its ranking, addressing why a docum…
NVIDIA released Cosmos 3 Edge, a 4-billion-parameter open world model designed to run on-device for robots and vision AI agents, enabling real-time reasoning and local action generation. The model is the smallest tier in…
NVIDIA's DGX Spark desktop GPU, with 128 GB unified memory and a 140W ARM64 processor, challenges API developers to shift from cloud-based AI consumption to local systems engineering. The device's shared memory architect…
NVIDIA released Alpamayo 2 Super, a 34B-parameter vision-language-action model for autonomous driving, under the OpenMDW-1.1 license, enabling commercial use. The model, built on a 32B Cosmos 3 Super Reasoner backbone an…
CopilotKit has released the Channels SDK, an MIT-licensed open source library that runs any AG-UI agent inside Slack and Microsoft Teams, with Discord and Google Chat planned. The SDK, installable via npm packages @copil…