
Meta's AI4AnimationPy Cuts Motion Capture Processing from Hours to Minutes
Meta open-sources AI4AnimationPy, a pure-Python framework that trains, runs, and visualizes neural character animation without a game engine
20 stories

Meta open-sources AI4AnimationPy, a pure-Python framework that trains, runs, and visualizes neural character animation without a game engine
A new paper from Stanford, Harvard, MIT, and Anthropic finally explains why bigger AI models outperform smaller ones — and it comes down to gradient interference, not just capacity.

Tencent's Hy-Memory plugin gives long-running agents a structured, evolving memory that outscores mem0 and Graphiti on every public benchmark

NVIDIA and Unitree unveil the first open humanoid robot reference design, giving academic labs a full-stack platform from simulation to real-world deployment
Runway joins NVIDIA's newly launched Cosmos Coalition as a founding member, a global push to build and open-source frontier world models for physical AI.

NVIDIA's Cosmos 3 merges reasoning, world simulation, and robot action generation into a single open model — a first for physical AI

NVIDIA's Nemotron 3 Ultra is a 550B open-weight MoE model that delivers 5x faster inference and 30% lower cost for agentic workloads, now the top US open-weight model by benchmark.

NVIDIA's 550B open-weights model tops US rankings with 5x faster inference, built for long-running AI agents — but China's Kimi K2.6 still leads globally

Nous Research's self-improving AI agent lands on Windows natively, no WSL required, as NVIDIA backs it at GTC Taipei

Researchers show merging K and V projections cuts KV cache by 50% with only 3.1% perplexity loss — and stacks with GQA/MQA for up to 96.9% reduction

A new geometric regularizer called Coh forces neural network features to tile data manifolds like brain cells, yielding provably interpretable latent spaces accepted at ICML 2026.

Researchers from the University of Michigan introduce MDA, a lightweight mixture-density head that eliminates 'flying point' artifacts in depth estimation without changing the backbone or adding inference overhead.
CG-CMARL combines coordination graphs with Lagrangian duality so a single trained model navigates the full safety-performance tradeoff for teams of up to 10 agents.
MOC rewires how LLM agents share information across multi-hop paths, cutting token bloat while boosting accuracy up to 6.77% on reasoning benchmarks

SIRI teaches LLM agents to mine and absorb reusable skills during training, then run lean at inference with no skill bank needed
CityTrajBench standardizes how we compare vehicle trajectory generators, revealing that no single model wins across all quality dimensions
A new audit of 1,603 NLP papers finds that who annotates, how they're trained, and what they're paid is routinely left out of published research
SeClaw auto-generates security test tasks from structured risk specs and evaluates agents inside Docker sandboxes, tracking every tool call along the way
POIROT turns a multi-agent system's own agents into its safety auditors, outperforming single-LLM evaluators with gains that scale with complexity
A new open-source pipeline turns a single smartphone video into 3D movement biomarkers for chronic pain patients, matching lab-grade motion capture at a fraction of the cost.