Lingfan Yu

Dr. Lingfan Yu

Annapurna Labs, Amazon, Cupertino, California, USA
orcid

Research Focus & Metrics

Academic Career Length

10Years (2017-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
23 April 2018
Verified
Conference Paper
Open access
PGPin Gao
Pin Gao
LYLingfan Yu
Lingfan Yu
YWYongwei Wu
Yongwei Wu
JLJinyang Li
Jinyang Li

14 October 2017
Verified
Conference Paper
Open access
CTCheng Tan
Cheng Tan
LYLingfan Yu
Lingfan Yu
JLJoshua B. Leners
Joshua B. Leners
MWMichael Walfish
Michael Walfish

19 November 2020 in arXiv
Verified
Preprint
Open access
PDF file
LYLingfan Yu
Lingfan Yu
JSJiajun Shen
Jiajun Shen
JLJinyang Li
Jinyang Li
ALAdam Lerer
Adam Lerer
Graph neural networks (GNNs) are a popular class of parametric model for learning over graph-structured data. Recent work has argued that GNNs primarily use the graph for feature smoothing, and have shown competitive results on benchmark tasks by simply operating on graph-smoothed node features, rather than using end-to-end learned feature hierarchies that are challenging to scale to large graphs. In this work, we ask whether these results can be extended to heterogeneous graphs, which encode multiple types of relationship between different entities. We propose Neighbor Averaging over Relation Subgraphs (NARS), which trains a classifier on neighbor-averaged features for randomly-sampled subgraphs of the "metagraph" of relations. We describe optimizations to allow these sets of node features to be computed in a memory-efficient way, both at training and inference time. NARS achieves a new state of the art accuracy on several benchmark datasets, outperforming more expensive GNN-based methods


26 April 2026
Verified
Conference Paper
Open access
YZYuan Zhou
Yuan Zhou
SXShaojie Xiang
Shaojie Xiang
LYLingfan Yu
Lingfan Yu
ZSZhenyu Song
Zhenyu Song
CMCharith Mendis
Charith Mendis
YWYida Wang
Yida Wang
Modern large language models rely on attention mechanisms that attend to all tokens in a sequence, resulting in quadratic computational complexity that limits scalability. While sparse attention reduces compute and memory requirements by attending to only important tokens, implementing these techniques presents significant challenges due to the complexity of combining static and dynamic sparse patterns and optimizing key-value (KV) cache management. To address these challenges, we present SAS, a sparse attention synthesizer that automatically generates performant sparse attention kernels for large language model inference. SAS introduces a set of primitives that effectively encapsulate both static and dynamic sparse attention mechanisms, enabling users to compose complex attention patterns through logic operators and declarative functions. The system employs a geometric-based pattern analyzer to optimize for KV caching by determining minimal cache sizes and automatically generating cache management functions. Supporting both Nvidia GPU and AWS Trainium backends, SAS demonstrates significant performance improvements: 1.10–1.22× speedup for context encoding and 2.68–2.80× speedup for token generation over FlexAttention, a state-of-the-art flexible attention kernel synthesis tool, on GPUs, and 1.41–6.49× speedup for context encoding and 1.39–10.87× speedup for token generation over optimized dense attention on Trainium.