小红书【Ace顶尖实习生】面向大模型推理提速的CoT压缩算法研究
实习兼职机器学习平台地点：北京 | 上海 | 杭州状态：招聘
任职要求
1、不限年级，本科及以上在读，计算机/人工智能/软件工程等相关专业优先； 2、熟悉Linux/Unix平台上的C++编程，熟悉网络编程-多线程编程，有良好的编程习惯； 3、熟悉其中一种主流的深度学习训练或推理框架（TensorFlow / PyTorch / Onnx / TensorRT等）的原理和实现者优先； 4、有扎实的专业基础知识，熟悉常用的数据结构和算法，对计算机系统结构-网络-操作系统等专业知识有深刻认知； 5、良好的沟通协作能力，责任心强，积极主动，能和团队一起探索新技术，推进技术进步。
工作职责
随着大型语言模型（LLMs）的快速发展，其在复杂任务中的推理效率问题日益凸显。本课题聚焦于LLMs的推理加速，旨在研究高效的Chain-of-Thought（CoT）压缩算法，以优化模型的推理过程，减少计算开销并提高响应速度，同时保持推理的准确性；同时，课题将深入分析现有LLMs的推理机制，探索如何通过算法创新来实现CoT的高效压缩。 具体研究内容包括但不限于：基于模型结构进行优化、基于推理过程进行优化、基于Prompt进行优化、以及基于数据驱动的压缩策略等。通过本课题的研究，期望能够为LLMs的高效推理提供新的理论和技术支持，推动其在更多实际场景中的广泛应用。
包括英文材料
Linux+
https://ryanstutorials.net/linuxtutorial/
Ok, so you want to learn how to use the Bash command line interface (terminal) on Unix/Linux.
https://ubuntu.com/tutorials/command-line-for-beginners
The Linux command line is a text interface to your computer.
https://www.youtube.com/watch?v=6WatcfENsOU
In this Linux crash course, you will learn the fundamental skills and tools you need to become a proficient Linux system administrator.
https://www.youtube.com/watch?v=v392lEyM29A
Never fear the command line again, make it fear you.
https://www.youtube.com/watch?v=ZtqBQ68cfJc
Unix+
[英文] The UNIX® Standard
https://www.opengroup.org/membership/forums/platform/unix
https://www.youtube.com/watch?v=IrDUcdpPmdI
UNIX is an operating system which was first developed in the 1970s, and has been under constant development ever since.
C+++
https://www.learncpp.com/
LearnCpp.com is a free website devoted to teaching you how to program in modern C++.
https://www.youtube.com/watch?v=ZzaPdXTrSb8
网络编程+
https://www.youtube.com/watch?v=2HrYIl6GpYg
I will make a simple HTTP web server with the C Programming Language.
https://www.youtube.com/watch?v=8z6okCgdREo
This tutorial is for Gophers who have written a command line or an API application, but have little to no experience in lower-level concepts like reading and writing to sockets, working with channels, and managing multiple goroutines.
https://www.youtube.com/watch?v=bdIiTxtMaKA&list=PL9IEJIKnBJjH_zM5LnovnoaKlXML5qh17
https://www.youtube.com/watch?v=bzja9fQWzdA
Implement the ubiquitous TCP protocol that underlies much of the traffic on the internet!
[英文] 📺Network Programming with Python Course (build a port scanner, mailing client, chat room, DDOS)
https://www.youtube.com/watch?v=FGdiSJakIS4
Learn network programming in Python by building four projects. You will learn to build a mailing client, a DDOS script, a port scanner, and a TCP Chat Room.
https://www.youtube.com/watch?v=gntyAFoZp-E
https://www.youtube.com/watch?v=JiuouCJQzSQ
Explore the fundamentals of networking in Rust by building a simple TCP server.
https://www.youtube.com/watch?v=JRTLSxGf_6w
https://www.youtube.com/watch?v=sFizpxHkIlI
In this video we'll cover SOCKET PROGRAMMING in JAVA.
https://www.youtube.com/watch?v=sXW_sNGvqcU
多线程+
https://liaoxuefeng.com/books/java/threading/basic/index.html
和单线程相比，多线程编程的特点在于：多线程经常需要读写共享数据，并且需要同步。
https://www.youtube.com/watch?v=_uQgGS_VIXM&list=PLsc-VaxfZl4do3Etp_xQ0aQBoC-x5BIgJ
https://www.youtube.com/watch?v=IEEhzQoKtQU
https://www.youtube.com/watch?v=mTGdtC9f4EU&list=PLL8woMHwr36EDxjUoCzboZjedsnhLP1j4
https://www.youtube.com/watch?v=TPVH_coGAQs&list=PLk6CEY9XxSIAeK-EAh3hB4fgNvYkYmghp
https://www.youtube.com/watch?v=xPqnoB2hjjA
This video is an introduction to multithreading in modern C++.
https://www.youtube.com/watch?v=YKBwKy5PrpQ
Rust threading is easy to implement and improves the efficiency of your applications on multi-core systems!
编程规范+
[英文] Google Style Guides
https://google.github.io/styleguide/
Every major open-source project has its own style guide: a set of conventions (sometimes arbitrary) about how to write code for that project. It is much easier to understand a large codebase when all the code in it is in a consistent style.
深度学习+
https://d2l.ai/
Interactive deep learning book with code, math, and discussions.
TensorFlow+
https://www.youtube.com/watch?v=tpCFfeUEGs8
Ready to learn the fundamentals of TensorFlow and deep learning with Python? Well, you’ve come to the right place.
https://www.youtube.com/watch?v=ZUKz4125WNI
This part continues right where part one left off so get that Google Colab window open and get ready to write plenty more TensorFlow code.
PyTorch+
https://datawhalechina.github.io/thorough-pytorch/
PyTorch是利用深度学习进行数据科学研究的重要工具，在灵活性、可读性和性能上都具备相当的优势，近年来已成为学术界实现深度学习算法最常用的框架。
https://www.youtube.com/watch?v=V_xro1bcAuA
Learn PyTorch for deep learning in this comprehensive course for beginners. PyTorch is a machine learning framework written in Python.
TensorRT+
https://docs.nvidia.com/deeplearning/tensorrt/latest/getting-started/quick-start-guide.html
This TensorRT Quick Start Guide is a starting point for developers who want to try out the TensorRT SDK; specifically, it demonstrates how to quickly construct an application to run inference on a TensorRT engine.
数据结构+
https://www.youtube.com/watch?v=8hly31xKli0
In this course you will learn about algorithms and data structures, two of the fundamental topics in computer science.
https://www.youtube.com/watch?v=B31LgI4Y4DQ
Learn about data structures in this comprehensive course. We will be implementing these data structures in C or C++.
https://www.youtube.com/watch?v=CBYHwZcbD-s
Data Structures and Algorithms full course tutorial java
算法+
https://roadmap.sh/datastructures-and-algorithms
Step by step guide to learn Data Structures and Algorithms in 2025
https://www.hellointerview.com/learn/code
A visual guide to the most important patterns and approaches for the coding interview.
https://www.w3schools.com/dsa/
相关职位
实习策略算法
本课题的研究目标是打造行业领先的多语言能力强化的大语言模型，结合业务海量的多语言互译语料，以及平台生活化特色的笔记评论数据，利用数据合成、RL冷启训练、SFT和RLHF等技术，实现： 1、在小红书多语言大模型翻译场景取得领先效果； 2、多语言场景下，AI搜索的检索和生成技术。
更新于 2025-07-20
实习大模型
本课题的研究目标是开发面向大模型的可扩展监督对齐方法，通过探索更高效的人类反馈收集与利用机制、自动化偏好学习和评估框架，以及对齐传递技术，实现在有限人类监督资源下对大规模AI系统进行有效对齐。 项目将建立产品与研究的协同设计机制，确保对齐技术能够直接响应实际产品需求，通过从产品应用场景中收集真实用户反馈来迭代优化对齐方法，形成研究与产品互促共进的闭环，提升模型在安全性、价值观一致性和指令遵循能力等方面的表现。
更新于 2025-08-22
实习基础后端
本课题聚焦在多模态数据场景下如何高效的组织其向量索引数据，设计并实现配套的混合查询优化技术，能根据用户请求动态选择合理的索引组合，实现耗时、吞吐的合理平衡。 预期成果是能用单一向量数据库产品原生支持异构多模态数据的向量索引构建和混合查询，为多模态大模型场景提供高效的向量支持。
更新于 2025-07-19
实习大模型
本课题的研究目标是研发面向复杂场景的可拓展大规模强化学习系统。 包括不限于： 1、超大规模Reward System构建； 2、面向复杂场景构建“策略与反馈一体化”的递归自我增强方法，解决“AI超越人类”时的可拓展监督问题； 3、面向长程任务探索人机合作博弈的强化学习机制，实现模型在超长程复杂任务上的需求明确、自我规划与执行校验能力；让AI从被动完成指令的工具转变为主动推动任务进展的的协作者，实现目标对齐； 4、强化学习进程中的大模型可解释性、可理解性。
更新于 2025-08-22