TokenUnify: Scaling Up Autoregressive Pretraining for Neuron Segmentation Computer VisionSelf-Supervised Learning
ICCV · 2025
Yinda Chen*; Haoyuan Shi*; Xiaoyu Liu; Te Shi; Ruobing Zhang; Dong Liu; Zhiwei Xiong; Feng Wu

Code

Dataset

Weights

Autoregressive pretraining for volumetric neuron segmentation, combining three complementary prediction objectives with long-sequence modeling and an analysis of prediction-error accumulation.

Self-Supervised Neuron Segmentation with Multi-Agent Reinforcement Learning Computer VisionSelf-Supervised Learning
IJCAI · 2023
Yinda Chen; Wei Huang; Shenglong Zhou; Qi Chen; Zhiwei Xiong

Code	Pretraining data

Decision-based masked image modeling that learns masking policies through multi-agent reinforcement learning for neuron segmentation.

Learning Multiscale Consistency for Self-Supervised Electron Microscopy Instance Segmentation Computer VisionPretraining Methods
ICASSP · 2024
Yinda Chen; Wei Huang; Xiaoyu Liu; Shiyu Deng; Qi Chen; Zhiwei Xiong

Code

Self-supervised pretraining that encourages multiscale consistency in volumetric electron microscopy representations for neuron instance segmentation.

Learned Image Coding with Generative Reference of Conditional Latents Image CompressionGenerative Models
IEEE TPAMI · Accepted October 2026
Siqi Wu*; Yinda Chen*; Weiming Chen; Dong Liu; K. C. Ho; Zhihai He

PDF

Code

GRCL uses retrieved or generated images as references for conditional latent coding. The framework combines reference generation, latent matching and synthesis, and theoretical analysis of robustness to reference perturbations.

Conditional Latent Coding with Learnable Synthesized Reference for Deep Image Compression Image Compression
AAAI · 2025
Siqi Wu*; Yinda Chen*; Dong Liu; Zhihai He

PDF

Code

Weights

Learnable reference synthesis from an external image dictionary improves conditional latent coding for learned image compression.

GTGM: Generative Text-Guided 3D Vision-Language Pretraining for Medical Image Segmentation Vision-LanguageMedical Imaging
ICCV Workshop on VLM3D · 2025
Yinda Chen*; Che Liu*; Wei Huang; Xiaoyu Liu; Haoyuan Shi; Sibo Cheng; Rossella Arcucci; Zhiwei Xiong

Code

Uses generated textual descriptions and negative-free contrastive learning to pretrain 3D medical image representations without paired medical text; evaluated through downstream segmentation on 10 CT/MRI datasets.

BIMCV-R: A Landmark Dataset for 3D CT Text-Image Retrieval Vision-LanguageMultimodal Learning
MICCAI · 2024
Yinda Chen; Che Liu; Xiaoyu Liu; Rossella Arcucci; Zhiwei Xiong

Dataset

Code

A dataset of 8,069 3D CT volumes with paired radiology reports, together with the MedFinder framework for bidirectional medical image–text retrieval.

MaskFactory: Towards High-quality Synthetic Data Generation for Dichotomous Image Segmentation Multimodal Learning
NeurIPS · 2024
Haotian Qian*; Yinda Chen*; Shengtao Lou; Fahad Shahbaz Khan; Xiaogang Jin; Deng-Ping Fan

Project

Code

A two-stage framework for generating paired images and masks for dichotomous image segmentation, expanding training data while preserving image–mask consistency.

EMPOWER: Evolutionary Medical Prompt Optimization With Reinforcement Learning Vision-LanguageMultimodal Learning
IEEE JBHI · 2025
Yinda Chen*; Yangfan He*; Jing Yang; Dapeng Zhang; Zhenlong Yuan; Muhammad Attique Khan; Jamel Baili; Por Lip Yee

Paper

Combines evolutionary prompt optimization with reinforcement learning and multi-dimensional evaluation for medical language tasks.

IQuest-Q1: Advancing Foundation Capabilities for Agentic CLI Systems Agentic ModelsTool Use
Ubiquant (九坤投资) · October 2026
Yinda Chen, co-author

Report

Project

Code

Models

A foundation model for command-line interaction, coding, and multi-step tool use. Contributed during my research internship at Ubiquant (九坤投资).

SAIL: Scientific Agentic Intelligence via a Science-Aware Loop Scientific AgentsAgentic Models
Ubiquant (九坤投资) · October 2026
SAIL Model Team · Yinda Chen, core contributor

Report

A scientific-agent research project for literature research, scientific coding, and multi-step investigations, developed through a science-aware improvement loop.

Pelican-Unified 1.0: A Unified Embodied Intelligence Model (UEI) for Understanding, Reasoning, Imagination and Action Embodied IntelligenceWorld ModelsUnified Models
X-Humanoid, WFM System Group · May 2026
Pelican-Unified Team · Yinda Chen, core contributor

My contribution: unified vision-language modeling and action capability.

A unified embodied model that combines scene and instruction understanding, task reasoning, and joint prediction of future video and robot actions.

Pelican-VLA 0.5: Attending Before Acting Benefits Generalization Vision-Language-ActionEmbodied IntelligenceWorld Models
X-Humanoid, WFM System Group · July 2026
Pelican-VLA Team · Yinda Chen, contributor

Report

Code

Models

The VLA component of Pelican-Unified connects vision-language understanding, future-frame generation, and action prediction through learnable bottleneck tokens. The work studies task-relevant attention and manipulation generalization.