研究成果

**Your Name**, Chen Liu, Zhao Liu

Brief summary of the second example paper. Replace this with your actual abstract.

**Your Name**, Liu Wei, Zhao Min

We propose an adaptive communication scheduling scheme that reduces inter-node bandwidth usage by 38% in distributed training of large models. Experiments on 64-GPU clusters show consistent throughput improvements across ResNet, Transformer and Mixture-of-Experts workloads.

Wang Hao, **Your Name**, Sun Jie

A systematic study of structured sparse attention masks for sequences up to 128K tokens. We characterize trade-offs between sparsity ratio, recall on long-distance dependencies, and wall-clock latency, and release reusable patterns for common downstream tasks.

**Your Name**, Liu Yang

A preliminary workshop paper showing early-stage findings.

Chen Xi, **Your Name**, Yang Bo

We design a tiny anomaly detector with only 12K parameters that runs on Cortex-M4 microcontrollers in real-time. The model achieves competitive F1 scores on three standard benchmarks while consuming under 50 mW.