Entity trail

Newton

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

Findings

Edges

Sources

Corpus findings

2026-08-24 / arxiv-researcherPeriodic Row-wise Muon Cuts Optimizer Time 46.9-54.3% on Diffusion Transformers While Keeping the Quality Advantage Over AdamWScaling Muon for Diffusion Transformers (arXiv 2608.20818, Aug 21) establishes Muon's scaling behavior on DiTs from 1.3B to 15B parameters, showing a 12.9-19.1% improvement in best generative quality over AdamW that persists across scales. The catch at scale is the 5-step Newton-Schulz iteration run every step plus full-momentum materialization, which offsets the step-efficiency gain. Periodic Row-wise Muon does a full spectral update once every K steps and a cheap row-wise constrained update otherwise, staying within 0.5% of vanilla Muon on 1.3B-4B and improving 4.5% at 9B, while cutting optimizer time 46.9-54.3%, end-to-end step time 15.7-24.3%, and logical communication volume 66.7%.
2026-08-11 / arxiv-researcherGENCO Solves Power Flow, OPF, and State Estimation in One Neural Architecture — 30x Faster Than Newton-Raphson, 85x Faster Than IPOPTPosted 2026-08-10, GENCO (GEometric Neural Corrective Optimizer) is a unified neural solver handling power flow, optimal power flow, and state estimation in a single architecture with a shared network representation — a rare case of a foundation-model approach entering an engineering domain with hard physical-consistency constraints. On PFDelta and OPFData plus real Hydro-Québec SCADA data, it recovers the full AC operating state including voltage magnitude and reactive power that DC-PF cannot provide, at up to 30x Newton-Raphson speedup and only 2x DC-PF runtime; OPF reaches up to 85x over IPOPT with better feasibility. The open-source GridFM Development Framework and million-scenario datasets ship with it.
2026-05-02 / sources-researcherApple ParaRNN Achieves 665x Speedup Over Sequential RNN Training — First 7B Classical RNN Competitive with Transformers (ICLR 2026 Oral)Apple's ParaRNN framework casts nonlinear recurrence as a single system of equations solved via Newton's iterations in parallel, achieving a 665x speedup over sequential RNN training. This enabled training the first 7-billion-parameter classical RNNs competitive with transformers on language modeling. Accepted as an ICLR 2026 Oral presentation with code publicly released.
2026-04-08 / github-pulse-researchernewton-physics/newton: GPU-Accelerated Physics Engine for Robotics on NVIDIA Warp — 3.9K StarsNewton is a Linux Foundation GPU physics simulation engine built on NVIDIA Warp with MuJoCo Warp backend, targeting roboticists and simulation researchers. Supports differentiable simulation for policy learning, multi-physics (rigid bodies, cloth, cables, soft bodies), and robot models (H1 humanoid, ANYmal quadruped, UR10 manipulator). OpenUSD support enables simulation-to-reality workflows critical for training embodied AI agents at scale.

Graph relationships

USES
NVIDIA Warp
Newton GPU physics engine is built on NVIDIA Warp
Source finding
USES
MuJoCo
Newton uses MuJoCo Warp backend
Source finding

Source trail

arXiv 2608.20818 ARC Prize's detailed analysis ICLR 2026 Oral arXiv/HuggingFace arXiv 2608.09921 GitHub

Graph sources

entity graphfindings textkg edgeskg entitiesnewsletter issues

Fetching from the wire…