Sparse data is optimally reused 15–20 times, leading to the proposal of a scaling law that reflects this.
Apple proposed LINK, a data intervention technique leveraging bilingual lexicons, to significantly improve multilingual LLM training efficiency.
AIUC-1 is the world's first independent standard for verifying the security and safety of AI agents.
OpenAI has paused reinforcement learning for its frontier models and strengthened safety measures after identifying cybersecurity risks.
A new technique leveraging parallel ANE and GPU execution on Apple Silicon improves LLM Prefill speed by up to 50%.
SkillEvaluator, developed in collaboration with NVIDIA, has been introduced to ClawHub to quantitatively demonstrate performance improvements in agent skills.
AI model marketplace OpenRouter has been acquired by Stripe and will be integrated.
Superwhisper's 600M parameter LLM, S1-mini, runs locally in the browser via WebGPU.
Inco AI released DFlash 2, a parallel inference acceleration technology that improves throughput by 16–25%.
OpenAI has applied Zero Data Retention to frontier models and released Private Safety Processing in preview.
Google announced it will provide college students worldwide with a free one-year AI plan and dedicated learning tools.
Unsloth has released Dynamic v3.0 quantized models for Qwen3.8-27B, demonstrating improved accuracy.
This guide explains how to centrally manage pull requests and issues using the 'My work' panel in the GitHub Copilot app.
Waymo has begun offering in-car AI assistant features by integrating Gemini into its proprietary Ojai vehicle.
Airbnb provided criteria for deciding whether to refit, respecify, or hold models during retraining.
A study highlights the 'unfaithfulness' issue where LLM CoT reasoning does not align with the actual internal process.
Cloudflare re-evaluated the risk of remote Spectre attacks in the Workers environment and strengthened its defense mechanisms.
AntLing has open-sourced six Ling-3.0-based models applying WSM technology.
Terence Tao proposed that the mathematical community redefine its goals and values as AI performs research-level mathematical tasks at ICM 2026.
Ornith-1.5 achieves open-source SOTA through self-improvement via self-generated tasks and reinforcement learning
AI inference chip developer Etched raised $700 million, more than doubling its valuation in just one month.
Amidst backlash against data centers in the US, 183 cities have implemented regulations, driving a trend of migrating AI infrastructure to sea and space.
Liquid AI has released the LFM2.5 Q4_0 models, which minimize quality degradation through Quantization-Aware Distillation (QAD).
Tencent ARC released SCoPE, a video generation model that controls camera trajectories.
AI Briefing