Longyin Wen

Research Manager@Bytedance Inc.·美国·OpenReview

研究方向

Multi-modality Understanding · Vision and Language · Image and Vision Editing · Self-Supervised Learning · Deep Learning · Object Detection · Video Segmentation · Poes Estimation · Object Tracking

6.5

CyberV: A Cybernetic Framework for Enhancing Logical Reasoning in Video Understanding

ICLR 2026Withdrawn

6.0

Multi-Reward as Condition for Instruction-based Image Editing

ICLR 2025Poster

5.8

Rethinking Homogeneity of Vision and Text Tokens in Large Vision-and-Language Models

ICLR 2025Rejected

通讯

5.3

DiffLM: Controllable Synthetic Data Generation via Diffusion Language Models

ICLR 2025Rejected

通讯

3.7

AIPO: Agreement-Aware Iterative Preference Optimization for Length Exploitation Mitigation

合作者 (20)

Longyin Wen

Referring Layer Decomposition

TGT: Text-Grounded Trajectories for Locally Controlled Video Generation

Where do Large Vision-Language Models Look at when Answering Questions?

CyberV: A Cybernetic Framework for Enhancing Logical Reasoning in Video Understanding

Multi-Reward as Condition for Instruction-based Image Editing

Rethinking Homogeneity of Vision and Text Tokens in Large Vision-and-Language Models

DiffLM: Controllable Synthetic Data Generation via Diffusion Language Models

AIPO: Agreement-Aware Iterative Preference Optimization for Length Exploitation Mitigation