Faster and Better LLMs via Latency-Aware Test-Time Scaling
EMNLP Empirical Methods in Natural Language Processing (Findings), 2025
Publications
Efficient Inference
MoESD: Unveil Speculative Decoding's Potential for Accelerating Sparse MoE
NeurIPS🏆 Spotlight Annual Conference on Neural Information Processing Systems, 2025
Long-Context Modeling
Revisiting the Integration of Convolution and Attention for Vision Backbone
NeurIPS Annual Conference on Neural Information Processing Systems, 2024
RelayAttention for Efficient Large Language Model Serving with Long System Prompts
ACL Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), 2024
BiFormer: Vision Transformer with Bi-Level Routing Attention
CVPR IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023
Other
openPangu DeepDiver-V2: Multi-Agent Learning for Deep Information Seeking
preprint Technical Report, 2025
Think Before You Talk: Enhancing Meaningful Dialogue Generation in Full-Duplex Speech Language Models with Planning-Inspired Text Guidance
preprint arXiv preprint arXiv:2508.07375, 2025
Analytic-Splatting: Anti-Aliased 3D Gaussian Splatting via Analytic Integration
ECCV🏆 Oral European Conference on Computer Vision, 2024
Inverse Rendering of Glossy Objects via the Neural Plenoptic Function and Radiance Fields
CVPR IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024
Towards Self-Adaptive Pseudo-Label Filtering for Semi-Supervised Learning
preprint arXiv preprint arXiv:2309.09774, 2023
Neural Preset for Color Style Transfer
CVPR IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023
Harmonizer: Learning to perform white-box image and video harmonization
ECCV European Conference on Computer Vision, 2022
Mitigating intensity bias in shadow detection via feature decomposition and reweighting
ICCV International Conference on Computer Vision, 2021