- Indonesia
- https://ruwwww.github.io
Stars
“Classical Energies Learn to Reason 1,000× Faster than Looped Models” (CRAM) by Benjamin Hoover, Hugh Blayney, and James C. R. Whittington: arXiv:2611.17309v1 (https://arxiv.org/pdf/2611.17309v1)
The Most Advanced MCP Server for Unity Editor with execute_code, prompts/resources, input simulation, screenshots, and play mode automation.
Formal verification workflow and skill stack for coding agents (TLA+, Z3, Apalache, TLAPS)
A full-stack, open-source NPU for LLM inference.
OPAL: a reactive optimizer. A coding agent rewrites the optimizer while your model trains.
🔥🔥 The extremely fast inference engine for decision model
FrontierVideo: Benchmarking Coding Agents for Professional Video Production
InstantMesh: Efficient 3D Mesh Generation from a Single Image with Sparse-view Large Reconstruction Models
NitroFusion2: Bringing Adversarial Loss to the Text Stream of Multimodal Diffusion Transformers.
Code for Latent Sparse Attention: Towards Subquadratic Ultra-Long Context Attention
Visual image editing, Anima regional prompting, and MOSS-Image layout prompting for classic ComfyUI.
HairDeck: Animation-Ready Stylized Hair Cards from a Single Image
Towards a More Autonomous Agent System
UEFI pre-driver NVIDIA GPU unlocker for Windows Linux Hackintosh and all EFI-bootable system (Rust): SEC2 scrubber -> FWSEC/WPR2 -> signed Booter -> V67 four-write payload, read-back verified. Veri…
Thermodynamic Hypergraphical Model Library in JAX
[CVPR 2025 Highlight🌟] Official ComfyUI implementation of "HyperLoRA: Parameter-Efficient Adaptive Generation for Portrait Synthesis"
Official code for Intrinsic Distribution Matching (IDM).
Research code for Anima reference adapters: synthetic data generation, training (single ref, two-slot, style_transfer knob), held-out evaluation, experiment log
Reference-image character and style transfer for Anima in ComfyUI (offset-RoPE reference frame + LoRA)
Schedule-Free Optimization in PyTorch
Train LoRAs inside ComfyUI: image and video models, live previews, on NVIDIA or AMD.
Your GeForce runs AI at half speed. This fixes it for 20/30/40 series cards. Up to 2× FP16 / BF16 / TF32 tensor throughput
CroissAttn: FP4 attention forward and backward, with analysis of stochastic rounding, zero-sum softmax gradients, and sampled centering.
MEND: RL for Flow Models via Proximal Velocity Matching. Official code, public model weights and inference. arXiv:2610.05954.
Minecraft: Java Edition 26.3 rewritten from scratch in Rust, checked against vanilla with 281 differential harnesses. Get it with the Horizon launcher.