<aside> 💡 We are seeking cooperators to work on domain adaptation of large language models (LLMs) to efficiently connect LLMs to various fields😊. If you are interested in working with us, feel free to reach out to Zhi Wang or Duo Wu🤗.
Zhi Wang
Email: [[email protected]](<mailto:[email protected]>)
Homepage: [<http://zwang.inflexionlab.org>](<http://zwang.inflexionlab.org/>)
Duo Wu
Email: [[email protected]](<mailto:[email protected]>)
Homepage: [<https://duowuyms.github.io/>](<https://duowuyms.github.io/>)
</aside>
Internship Recruitment (实习生招生)
With billions of parameters pre-trained on massive data to absorb extensive knowledge, LLMs have demonstrated extraordinary capabilities in a myriad of NLP tasks. Inspired by their remarkable success, we are dedicated to expand the boundaries of their capabilities and explore their potential in other domains beyond NLP. However, the domain adaptation of LLMs encounters the following primary challenges:
To tackle the above challenges, our research focuses on designing advanced algorithms and frameworks to empower LLMs to efficiently solve problems across various domains, based on the cutting-edge techniques such as multimodal learning, transfer learning, prompt engineering and reinforcement learning.
Many networking tasks now employ deep learning (DL) to solve complex prediction and system optimization problems. However, current design philosophy of DL-based algorithms entails intensive engineering overhead due to the manual design of deep neural networks (DNNs) for different networking tasks. Besides, DNNs tend to achieve poor generalization performance on unseen data distributions/environments.
Motivated by the recent success of large language models (LLMs), for the first time, this work studies the LLM adaptation for networking to explore a more sustainable design philosophy. With the massive pre-trained knowledge and powerful inference ability, LLM can serve as the foundation model, and is expected to achieve “one model for all” with even better performance and stronger generalization for various tasks. In this paper, we present NetLLM, the first LLM adaptation framework that efficiently adapts LLMs to solve networking problems. NetLLM addresses many practical challenges in LLM adaptation, from how to process task-specific information with LLMs, to how to improve the efficiency of answer generation and acquiring domain knowledge for net working. Across three networking-related use cases- view port prediction (VP), adaptive bitrate streaming (ABR) and cluster job scheduling (CJS), we showcase the effectiveness of NetLLM in LLM adaptation for networking. Results show that the adapted LLM surpasses state-of-the-art algorithms by 10.1-36.6% for VP, 14.5-36.6% for ABR, 6.8-41.3% for CJS, and also achieves superior generalization performance.
Duo Wu, Xianda Wang, Yaqi Qiao, Zhi Wang, Junchen Jiang, Shuguang Cui, Fangxin Wang. NetLLM: Adapting Large Language Models for Networking. Accepted by SIGCOMM 2024 [CCF-A] [preprint]
<aside> 💡 We plan to make extensions to our work. Free free to contact us if you have any suggestions or would like to join us. 😀
</aside>
Tool augmentation has emerged as an effective approach for enhancing the capabilities of large language models (LLMs) by integrating them with external tools, such as knowledge bases, math calculators, and code interpreters. Benefiting from the flexibility of plug-and-play tools, this approach offers a promising way for LLMs to efficiently solve problems across diverse domains.
The primary challenge of tool augmentation lies in deriving optimal tool scheduling topologies that enable accurate generation of answers and responses, given a set of pre-defined tools and user queries. In this project, our objective is to exploit LLM itself as the planner for tool scheduling. We aim to design algorithms and systems that enable LLMs to generate high-quality tool scheduling topologies to address complex problems under specific resource constraints (e.g., latency budgets).