A decentralized agent network for building collaborative, LLM-powered agent-to-agent (A2A) systems.

"LightReasoner: Can Small Language Models Teach Large Language Models Reasoning?"

RepoMaster: The open-source AI agent that masters GitHub. It turns any code repository into a powerful tool, achieving a new level of autonomous task-solving. An open alternative to Claude-Code.

[BIRD-INTERACT] Re-imagines Text-to-SQL evaluation via lens of dynamic interactions.

INFTY Engine: An Optimization Toolkit to Support Continual AI

(ICML 2024) Spider: A Unified Framework for Context-dependent Concept Segmentation

[CVPR'25] Official PyTorch implementation of AvatarArtist: Open-Domain 4D Avatarization.

NeuralOM: Neural Ocean Model for Subseasonal-to-Seasonal Simulation

Dataset and evaluation code of ISDrama(ACM-MM 2025): Immersive Spatial Drama Generation through Multimodal Prompting

Official Repo of "RobustFlow: Towards Robust Agentic Workflow Generation"

Deciphering Oracle Bone Language with Diffusion Models (ACL 2024 Best Paper)

Zen7 Payment Agent is the first implementation project of DePA (Decentralized Payment Agent), pioneers next-generation intelligent payment infrastructure.

Dynamic human image animation with strong identity preservation, heterogeneous character driving, and controllable backgrounds.

UR2: Unify RAG and Reasoning through Reinforcement Learning

[NeurIPS 2024] AWT: Transferring Vision-Language Models via Augmentation, Weighting, and Transportation

[USENIX Security'25] THEMIS: Towards Practical Intellectual Property Protection for Post-Deployment On-Device Deep Learning Models

your finance bro Agent for trading and investing

UniVoice: Unifying Autoregressive ASR and Flow-Matching based TTS with Large Language Models

[CVPR 2025] Test-Time Domain Generalization via Universe Learning: A Multi-Graph Matching Approach for Medical Image Segmentation

A project aims to improve LLMs' pixel reasoning ability.

[IROS 2025] MarineGym: A High-Performance Reinforcement Learning Platform for Underwater Robotics

An adaptive sampling framework for Reinforce-style LLM post training.

The Official Implementation for ''Towards Zero-shot Point Cloud Anomaly Detection: A Multi-View Projection Framework''

Lightweight replication study of DeepSeek-R1-Zero. Interesting findings include "No Aha Moment", "Longer CoT ≠ Accuracy", and "Language Mixing in Instruct Models".

Open, reproducible benchmarks and practical recipes to reduce I/O bottlenecks and improve end-to-end performance in AI training and bulk inference.