Xiangyu Zhang

Principal Researcher@MEGVII Technology·中国·OpenReview

研究方向

large language models · large vision-language models · large multimodal models · representative learning · self-supervised learning · deep learning · neural networks · CNNs · computer vision · recognition · object detection

Xiangyu Zhang

Mixture-of-Experts Can Surpass Dense LLMs Under Strictly Equal Resource

NoiseAR: AutoRegressing Initial Noise Prior for Diffusion Models

MemoryVLA: Perceptual-Cognitive Memory in Vision-Language-Action Models for Robotic Manipulation

NextStep-1: Toward Autoregressive Image Generation with Continuous Tokens at Scale

Predictable Scale (Part II) --- Farseer: A Refined Scaling Law in LLMs

Open-Reasoner-Zero: An Open Source Approach to Scaling Up Reinforcement Learning on the Base Model

GUI Exploration Lab: Enhancing Screen Navigation in Agents via Multi-Turn Reinforcement Learning

Unhackable Temporal Reward for Scalable Video MLLMs

Perception-R1: Pioneering Perception Policy with Reinforcement Learning

Open Vision Reasoner: Transferring Linguistic Cognitive Behavior for Visual Reasoning

Perception in Reflection

DreamBench++: A Human-Aligned Benchmark for Personalized Image Generation

Vision Foundation Models as Effective Visual Tokenizers for Autoregressive Generation

Reconstructive Visual Instruction Tuning

Glad: A Streaming Scene Generator for Autonomous Driving

PerPO: Perceptual Preference Optimization via Discriminative Rewarding

Is a 3D-Tokenized LLM the Key to Reliable Autonomous Driving?

Continual LLaVA: Continual Instruction Tuning in Large Vision-Language Models

General OCR Theory: Towards OCR-2.0 via a Unified End-to-end Model

Continuous Semi-Implicit Models

Generalizable Dynamic Radiance Field in Egocentric View

Image Generation with Channel-wise Quantization

MOTRv3: Release-Fetch Supervision for End-to-End Multi-Object Tracking

PADriver: Towards Personalized Autonomous Driving