pengzhangzhi
Follow
Lists (4)
Starred repositories
498
results
for source starred repositories
Clear filter
An implementation of a small scale autoencoder and latent diffusion model
Generative model for protein binder design for protein and small molecule targets. Combines a pretrained flow-based generative model (built on La-Proteina) with inference-time optimization.
Agent skills for Manim to create 3Blue1Brown style animations.
A lightweight, AI-native training framework for large language models. Designed for fast iteration, reproducible experiments, and modular configuration across SFT, RLVR, and evaluation workflows.
Minimal PyTorch implementation of Planner Aware Path Learning (PAPL) for diffusion language models, from our ICLR 2026 paper.
6
Updated Mar 2, 2026
Official Code Repo for "Learning on the Manifold: Unlocking Standard Diffusion Transformers with Representation Encoders"
16
Updated Feb 10, 2026
Single-step image generation at 306 FPS. Drifting vs Diffusion head-to-head on CIFAR-10.
SONAR, a new multilingual and multimodal fixed-size sentence embedding space, with a full suite of speech and text encoders and decoders.
Large Concept Models: Language modeling in a sentence representation space
Towards Scalable Pre-training of Visual Tokenizers for Generation
Minimal Claude Code alternative. Single Python file, zero dependencies, ~250 lines.
A repo for transferable sampling of molecular systems
Accelerate TarFlow Sampling with GS-Jacobi Iteration
Central repository for biomolecular foundation models with shared trainers and pipeline components
Official implementation of the NeurIPS 24 paper of statistical flow matching (SFM) for discrete generation.
Official Code Repository for the paper "Continuous Diffusion Model for Language Modeling" (NeurIPS 2025).
Official implementation of "Continuous Autoregressive Language Models"
Geometric Byte Pair Encoding of Protein Structure (ICLR 2026)
A simple but powerful VSCode extension for visualizing protein structures using the Molstar framework.
Repository for the BoltzGen model