Researcher — paper.dinq.me

2026 顶会论文

作者位次 #2/5PPoPP 2026Oral· Parallel & Distributed Programming

BEEMS: Boosting Machine Vision Efficiency via Computation Graph-Based Memory Smoothing

Hanjing Shen · Fangxin Liu · Jian Liu · Li Jiang · Haibing Guan

多模态效率与系统检索与知识OpenReview

作者位次 #6/9PPoPP 2026Oral· Parallel & Distributed Programming

Accelerating Sparse Transformer Inference on GPU

Wenhao Dai · Haodong Deng · Mengfei Rong · Xinyu Yang · Hongyu Liu · Fangxin Liu · Hailong Yang · Qianwen Cao · Qingxiao Sun

效率与系统代码OpenReview

作者位次 #3/9HPCA 2026Oral· High-Performance Computer Architecture

ORANGE: Exploring Ockham's Razor for Neural Rendering by Accelerating 3DGS on NPUs with GEMM-Friendly Blending and Balanced Workloads

Haomin Li · Yun Liang · Fangxin Liu · Bowen Zhu · Zongwu Wang · Yu Feng · Liqiang Lu · Li Jiang · Haibing Guan

长上下文与架构OpenReview

作者位次 #5/11HPCA 2026Oral· High-Performance Computer Architecture

SPLATONIC: Architectural Support for 3D Gaussian Splatting SLAM via Sparse Processing

Xiaotong Huang · He Zhu · Tianrui Ma · Yuxiang Xiong · Fangxin Liu · Zhezhi He · Yiming Gan · Zihan Liu · Jingwen Leng · Yu Feng · Minyi Guo

长上下文与架构OpenReview

作者位次 #2/7ISCA 2026Oral· Computer Architecture

COSM: A Cooperative Scheduling Framework for Concurrent PIM and CPU Execution on Mobile Devices

Yilong Zhao · Fangxin Liu · Onur Mutlu · Mingyu Gao · Jian Liu · Haibing Guan · Li Jiang

长上下文与架构OpenReview

一作ISCA 2026Oral· Computer Architecture

STEP: Adaptive Spatio-Temporal Expert Prefetching for Low-Latency and Memory-Efficient MoE Inference

Fangxin Liu · Ning Yang · Zongwu Wang · Chenyang Guan · Haomin Li · Yu Feng · Liqiang Lu · Xiang Li · Siran Yang · Jiamang Wang · Lin Qu · Li Jiang · Haibing Guan

效率与系统检索与知识长上下文与架构OpenReview

作者位次 #2/11ISCA 2026Oral· Computer Architecture

Harmonia: A Unified Hierarchical Scheduling Framework for Sparse Matrix Multiplication

Jingkui Yang · Fangxin Liu · Xin Ju · Ning Yang · Chenyang Guan · Junjie Wang · Zongwu Wang · Mei Wen · Jian Liu · Li Jiang · Haibing Guan

安全长上下文与架构OpenReview

作者位次 #7/9MICRO 2026Oral· ML System Interconect

SNN-4-All: Breaking the Edge Fine-Tuning Barrier with Heterogeneous Sparsity Exploitation

Zhibai Huang · James Yen · Zhixiang Wei · Xingzi Yu · Chen Chen · Lingyun Yang · Fangxin Liu · Zhengwei Qi · Haibing Guan

强化学习与后训练效率与系统OpenReview

作者位次 #8/10MICRO 2026Oral· Accelerators for Robotics, Vision & Graph Workloads

DeGS: A Scalable 3DGS Architecture via Decoupled Workload Parsing and Reorganization

Minnan Pei · Gang Li · Zeyu Zhu · Siting Wang · Junwen Si · Zhuoran Song · Yu Feng · Fangxin Liu · Xiaoyao Liang · Jian Cheng

多模态长上下文与架构OpenReview

作者位次 #6/12MICRO 2026Oral· Vision-Language & Multimodal Model Acceleration

Deltoris: Enabling Real-time VLA Inference in Embodied AI via Bit-level Sparsity and Speculative Inference

Zheng Liu · Zeyu Guo · Zihan Liu · Anbang Wu · Han Zhao · Fangxin Liu · Zhezhi He · Yinhe Han · Jingwen Leng · Minyi Guo · Yiming Gan · Yu Feng

效率与系统多模态OpenReview

作者位次 #3/11MICRO 2026Oral· PIM & Near-Memory Long-Context LLM Serving

ULTRA: Bridging the VQ Gap in Anisotropic LLM Quantization via a \underline{U}nified \underline{L}UT-Based \underline{TR}ansformer \underline{A}rchitecture

Zongwu Wang · Zhongyi Tang · Fangxin Liu · Shiyuan Huang · Chenyang Guan · Haomin Li · Gang Li · Liqiang Lu · Mingyu Gao · Li Jiang · Haibing Guan

效率与系统检索与知识长上下文与架构OpenReview

一作MICRO 2026Oral· PIM & Near-Memory Long-Context LLM Serving

MOSAIC: Exploiting Structured Tolerance for Adaptive LLM Inference Mapping in Heterogeneous PIM Accelerators

Fangxin Liu · Jingkui Yang · Yilong Zhao · Haomin Li · Wen Mei · Xinran Liang · Li Jiang · Haibing Guan

效率与系统检索与知识长上下文与架构OpenReview

作者位次 #8/11MICRO 2026Oral· Quantum & Emerging Technologies

TensorMLD: A Tensorized and Compressed qLDPC Maximum-Likelihood Decoder for Accurate and Fast Detection of the Most Likely Logical Error

Liqiang Lu · Debin Xiang · Kaiwen Zhou · Boyu Jin · Chenning Tao · Wuwei Tian · anbangwu · Fangxin Liu · Jingwen Leng · Minyi Guo · Jianwei Yin

推理效率与系统OpenReview

作者位次 #2/8ACL 2026Poster· Natural Language Processing

Calibrated Speculative Decoding: Frequency-Guided Candidate Selection for Efficient Inference

Zhouxuwen · Fangxin Liu · Chao Wang · Xiao Zheng · Hao Zheng · Min He · Li Jiang · Haibing Guan

摘要

Speculative decoding accelerates autoregressive generation by letting draft tokens bypass full verification, but conventional frameworks suffer from frequent false rejections, particularly when draft models produce semantically correct but lexically divergent outputs. In this paper, we present Ca...

效率与系统推理评测与基准OpenReview 项目页

作者位次 #2/7AAAI 2026Poster· Artificial Intelligence

SpecQuant: Spectral Decomposition and Adaptive Truncation for Ultra-Low-Bit LLMs Quantization

Zhixiong Zhao · Fangxin Liu · Junjie Wang · Chenyang Guan · Zongwu Wang · Li Jiang · Haibing Guan

效率与系统OpenReview

Fangxin Liu

获取完整研究档案

联系方式

2026 顶会论文

BEEMS: Boosting Machine Vision Efficiency via Computation Graph-Based Memory Smoothing

Accelerating Sparse Transformer Inference on GPU

ORANGE: Exploring Ockham's Razor for Neural Rendering by Accelerating 3DGS on NPUs with GEMM-Friendly Blending and Balanced Workloads

SPLATONIC: Architectural Support for 3D Gaussian Splatting SLAM via Sparse Processing

COSM: A Cooperative Scheduling Framework for Concurrent PIM and CPU Execution on Mobile Devices

STEP: Adaptive Spatio-Temporal Expert Prefetching for Low-Latency and Memory-Efficient MoE Inference

Harmonia: A Unified Hierarchical Scheduling Framework for Sparse Matrix Multiplication

SNN-4-All: Breaking the Edge Fine-Tuning Barrier with Heterogeneous Sparsity Exploitation

DeGS: A Scalable 3DGS Architecture via Decoupled Workload Parsing and Reorganization

Deltoris: Enabling Real-time VLA Inference in Embodied AI via Bit-level Sparsity and Speculative Inference

ULTRA: Bridging the VQ Gap in Anisotropic LLM Quantization via a \underline{U}nified \underline{L}UT-Based \underline{TR}ansformer \underline{A}rchitecture

MOSAIC: Exploiting Structured Tolerance for Adaptive LLM Inference Mapping in Heterogeneous PIM Accelerators

TensorMLD: A Tensorized and Compressed qLDPC Maximum-Likelihood Decoder for Accurate and Fast Detection of the Most Likely Logical Error

Calibrated Speculative Decoding: Frequency-Guided Candidate Selection for Efficient Inference

SpecQuant: Spectral Decomposition and Adaptive Truncation for Ultra-Low-Bit LLMs Quantization

常见合作者