Jiajun Liang

papers

1,505

total citations

papers (15)

Decoupled Knowledge Distillation

CVPR 2022arXiv

773

citations

Implicit Identity Leakage: The Stumbling Block to Improving Deepfake Detection Generalization

CVPR 2023arXiv

201

citations

Improving Video Generation with Human Feedback

NEURIPS 2025arXiv

127

citations

Joint Token Pruning and Squeezing Towards More Aggressive Compression of Vision Transformers

CVPR 2023arXiv

110

citations

Explaining Deepfake Detection by Analysing Image Matching

ECCV 2022arXiv

citations

Boosting Semi-Supervised Learning by Exploiting All Unlabeled Data

CVPR 2023arXiv

citations

Dynamic MLP for Fine-Grained Image Classification by Leveraging Geographical and Temporal Information

CVPR 2022arXiv

citations

Discriminability-Transferability Trade-Off: An Information-Theoretic Perspective

ECCV 2022arXiv

citations

DOT: A Distillation-Oriented Trainer

ICCV 2023arXiv

citations

MegActor-Sigma: Unlocking Flexible Mixed-Modal Control in Portrait Animation with Diffusion Transformer

AAAI 2025

citations

LEDiT: Your Length-Extrapolatable Diffusion Transformer without Positional Encoding

NEURIPS 2025arXiv

citations

Sharp Impossibility Results for Hyper-graph Testing

NEURIPS 2021

citations

Jiajun Liang

papers (15)

Decoupled Knowledge Distillation

Implicit Identity Leakage: The Stumbling Block to Improving Deepfake Detection Generalization

Improving Video Generation with Human Feedback

Joint Token Pruning and Squeezing Towards More Aggressive Compression of Vision Transformers

Explaining Deepfake Detection by Analysing Image Matching

Boosting Semi-Supervised Learning by Exploiting All Unlabeled Data

Dynamic MLP for Fine-Grained Image Classification by Leveraging Geographical and Temporal Information

A Simple Baseline for Efficient Hand Mesh Reconstruction

Efficient One Pass Self-Distillation with Zipf’s Label Smoothing

Cumulative Spatial Knowledge Distillation for Vision Transformers

Discriminability-Transferability Trade-Off: An Information-Theoretic Perspective

DOT: A Distillation-Oriented Trainer

MegActor-Sigma: Unlocking Flexible Mixed-Modal Control in Portrait Animation with Diffusion Transformer

LEDiT: Your Length-Extrapolatable Diffusion Transformer without Positional Encoding

Sharp Impossibility Results for Hyper-graph Testing

papers (15)

Decoupled Knowledge Distillation

Implicit Identity Leakage: The Stumbling Block to Improving Deepfake Detection Generalization

Improving Video Generation with Human Feedback

Joint Token Pruning and Squeezing Towards More Aggressive Compression of Vision Transformers

Explaining Deepfake Detection by Analysing Image Matching

Boosting Semi-Supervised Learning by Exploiting All Unlabeled Data

Dynamic MLP for Fine-Grained Image Classification by Leveraging Geographical and Temporal Information

A Simple Baseline for Efficient Hand Mesh Reconstruction

Efficient One Pass Self-Distillation with Zipf’s Label Smoothing

Cumulative Spatial Knowledge Distillation for Vision Transformers

Discriminability-Transferability Trade-Off: An Information-Theoretic Perspective

DOT: A Distillation-Oriented Trainer

MegActor-Sigma: Unlocking Flexible Mixed-Modal Control in Portrait Animation with Diffusion Transformer

LEDiT: Your Length-Extrapolatable Diffusion Transformer without Positional Encoding

Sharp Impossibility Results for Hyper-graph Testing