A 33M parameter English embedding model is climbing Hugging Face trending charts, offering 384 dimensional vectors with surprisingly competitive MTEB scores.
Snowflake's open-source Arctic Embed L delivers 55.98 NDCG@10 on MTEB Retrieval, matching closed APIs from OpenAI and Cohere at a fraction of the size.
A new regularizer called DSReg peels apart the mixed latents inside JEPA-style world models, recovering each individual world variable with no decoder and no labels.
Zyphra researchers cut MoE all-to-all communication by up to 2.63x on AMD MI300X GPUs by exploiting predictable token routing patterns, with no changes to the model.
A patched serving stack runs the 27B Ternary Bonsai 2 at the full 262k context window on a single 12 GB consumer GPU, with agentic tool-calling finally fixed.
Nous Research launches an agentic leaderboard that scores frontier models on four task suites inside the Hermes Agent harness, reporting both accuracy and dollar cost per task.
sparkDash is an open-source React and Express dashboard that streams live GPU, memory, and LLM metrics from any number of NVIDIA DGX Spark machines in one browser tab.
The former Amazon Alexa and Nova chief takes the helm at the Hyundai-owned robotics firm as it pivots from research darling to industrial Physical AI vendor.
Google turns Gemini Live into a conversational sighted guide for Android, trained with Aira interpreters across tens of thousands of hours of real-world scenes.
A new study finds no principled hyperparameter scaling rule consistently preserves optimizer performance across batch sizes, undermining single-batch benchmarks.