Distributed Computing

Bangjun Xiao, Yihao Zhao, Xiangwei Deng et al.

This work addresses resource management inefficiencies for cloud-based agentic RL systems, offering significant performance gains and cost savings, though it is incremental as it builds on existing frameworks with a novel orchestration approach.

Shubham Agarwal, Alexander Krentsel, Shu Liu et al.

For developers of safety-critical distributed systems, IDS dramatically reduces the effort and cost of formal verification, which previously required months to years of expert work.

Chen Zhou, Haoran Sun, Hedan Yang et al.

This work addresses infrastructure bottlenecks for researchers and developers in embodied AI, though it is incremental as it builds on existing frameworks like LeRobot.

Xinyi Hu, Yuhao Shen, Baolin Zhang et al.

This addresses the bottleneck of verification compute in production-grade LLM serving, offering a novel solution for high-concurrency scenarios.

Yukang Chen, Luozhou Wang, Wei Huang et al.

This work addresses speed and memory bottlenecks in long video generation for practitioners, offering a practical system-level solution.

Zhongzhu Zhou, Donglin Zhuang, Jisen Li et al.

This work addresses the challenge of accurate and deployable INT2 KV cache quantization for long-context LLM serving, offering a practical solution that integrates with existing frameworks.

You-Liang Huang, Xinhao Huang, Chengxi Liao et al.

For researchers and engineers scaling LLMs via decomposition, DeInfer provides a practical solution to a critical performance bottleneck.

Xinquan Chen, Zhenyun Yin, Shan He et al.

For researchers building autonomous agents, Safactory offers a unified framework for systematic risk discovery and continuous improvement, but the contribution is primarily architectural and lacks empirical validation.

Jae-Won Chung, Jeff J. Ma, Jisang Ahn et al.

This addresses the problem of efficient and scalable serving for complex multimodal models, which is incremental as it builds on existing distributed systems and Kubernetes.

Ruibo Fan, Xiangrui Yu, Xinglin Pan et al.

This addresses the problem of slow and memory-intensive LLM serving for AI practitioners, offering a novel co-designed solution that provides both compression and acceleration.

Yuhao Shen, Junyi Shen, Quan Kong et al.

This addresses inference latency issues for users of large language models, representing a significant but incremental improvement over existing speculative decoding methods.

Keisuke Kamahori, Shihang Li, Simon Peter et al.

This work challenges the paradigm of general-purpose LLM serving stacks by proposing generation-time specialization, which could benefit system builders and researchers dealing with diverse model architectures, workloads, and hardware.

Insu Jang, Runyu Lu, Nikhil Bansal et al.

For researchers and engineers training large multimodal models, Cornstarch provides a more efficient distributed training approach tailored to the heterogeneity of MLLMs.

Zengyi Qin, Jinyuan Chen, Yunze Man et al.

This addresses the resource-intensive infrastructure needed for computer use agent research, offering a scalable solution.

Xiulong Yuan, Hongqing Chen, Jiaxuan Peng et al.

For practitioners training complex multi-component LLMs, Maestro provides a practical framework that significantly improves GPU utilization and throughput over monolithic approaches.

Yadi Cao, Sicheng Lai, Jiahe Huang et al.

This addresses the gap in cost-aware evaluation for physics simulations, providing a practical benchmark for researchers and practitioners, though it is incremental in extending existing benchmarking approaches.

Muntaka Ibnath, Mohammadreza Rezvani, Daniel Wong

This addresses the complexity and overhead of instrumenting applications for QoS feedback in system management, though it appears incremental as an eBPF-based tool for observability.

Shuo Yang, Haocheng Xi, Yilong Zhao et al. · tsinghua

This work enables k-means as an efficient online primitive for modern AI systems, addressing a domain-specific bottleneck in GPU workloads.

Nina Wiedemann, Quentin Leboutet, Michael Paulitsch et al.

This addresses the problem of efficient GPU kernel optimization for developers and researchers, offering a novel method that improves performance over existing approaches.

Wei Gao, Yuheng Zhao, Dilxat Muhtar et al.

For LLM post-training systems, this addresses the inefficiency of static GPU provisioning for agentic RL rollouts.