Systems & Control

Andrew Lanpouthakoun, Aryaman Arora, Zhengxuan Wu et al.

For serving personalized LLMs with multiple adapters, PreFT offers a better accuracy-throughput tradeoff than existing PEFT methods.

Nengneng Yu, Sixian Xiong, Yibo Zhao et al.

Provides a practical systems primitive for model-internal observability in LLM inference, addressing the need for timely internal state access in inference-time workloads.

Elad Hazan, Karan Singh · princeton

This addresses the problem of robust control under adversarial conditions for researchers and practitioners in control theory and reinforcement learning, representing a foundational shift rather than an incremental improvement.

Thomas T. Zhang, Alok Shah, Yifei Zhang et al.

For practitioners training neural networks with one-step prediction losses but deploying them in multi-step rollouts, DoPr offers a drop-in optimization intervention to mitigate error accumulation and improve downstream metrics.

Shan Yu, Junyi Shu, Yuanjiang Ni et al.

For developers of multi-agent LLM applications, Pythia addresses inefficiencies in existing serving systems by leveraging predictable agent behavior.

Melissa Z. Pan, Negar Arabzadeh, Mathew Jacob et al.

For practitioners deploying retrieval agents, BRANE offers a practical per-query optimization method that outperforms static tuning and existing routing baselines.

Saurabh Bagchi, Hyunseung Kim, Tarek Abdelzaher et al.

For researchers and practitioners in CPS, this survey offers a structured overview of resilience challenges and solutions, but it is an incremental synthesis of existing work rather than a novel contribution.

Jiaqian Yang, Romulo Aparecido, Eric Sillekens et al.

This work demonstrates record bi-directional data transmission over hollow-core fiber, addressing the need for higher capacity in optical communications.

Sixu Li, Deepak Prakash Kumar, Swaroop Darbha et al.

Provides a theoretical and practical solution for time-optimal path planning on spheres, applicable to satellite attitude control and spherical robots, but is incremental as it extends known Reeds-Shepp results to a spherical domain.

Kalyan Cherukuri, Lav R. Varshney

This work addresses the issue of factual inaccuracies in LLM outputs for users relying on reliable text generation, though it is incremental in building on existing dynamical systems frameworks.

Boyuan Wang, Zhenyuan Zhang, Zhiqin Yang et al.

For robot learning from real-world data, PhysClaw-0 reduces human oversight cost per distinct failure mode rather than per episode, making large-scale autonomous data collection more practical.

Abdullah Bahamdan, Emma Pajak, John D. Hedengren et al.

This addresses a major bottleneck in process systems engineering by automating flowsheet generation, though it appears incremental relative to existing diagram-understanding and text-to-simulation methods.

Xingyu Feng, Chang Sun, Yuzhu Wang et al.

It addresses battery management for mobile device users by offering a safe, adaptive solution, though it is incremental as it builds on LLM capabilities for a specific domain.

Xihang Yu, Rajat Talak, Lorenzo Shaikewitz et al.

For robotics and simulation-based planning, this work provides a method to obtain physically plausible multi-object reconstructions, enabling more reliable digital twins for contact-rich tasks.

Yuanyi Wang, Yanggan Gu, Su Lu et al.

For practitioners merging large LLMs, MergePipe provides a budget-aware method to drastically reduce I/O costs while maintaining merge quality.

Yuhan Chen, Tao Liu, Jie Huang

This solves a long-standing challenge in control theory for multi-agent systems, enabling applications like robotics and autonomous vehicles with unstable dynamics, though it is a foundational breakthrough rather than incremental.

Xiucheng Wang, Yuhao Pan, Nan Cheng

It provides a comprehensive guide for researchers and engineers in wireless communications, though it is incremental as a survey rather than presenting new results.

Xiaojing Chen, Haiqi Yu, Wei Ni et al.

It addresses energy challenges for Agentic AI systems in mobile edge computing and wireless networks, but is incremental as it synthesizes existing knowledge into a survey.

Jihoon Hong, Alice Chan, Qiyue Dai et al.

For practitioners deploying text-to-video models, this provides a minimally invasive steering method that avoids oversteering and content degradation.

Emmanuel O. Badmus, Amritanshu Pandey

It addresses reliability issues in agentic AI for engineering workflows, offering incremental improvements in a domain-specific context.