Jianfeng Cai (蔡建峰)
Master Student, supervised by Prof. Houqiang Li and Wengang Zhou, Microsoft Key Laboratory of Multimedia Computing and Communication Lab, Department of Electronic Engineering and Information Science, University of Science and Technology of China
Currently, my research interests focus on Large Language Models and Agents. If you have any things or questions to discuss, do not hesitate to contact me.
[Wechat] [QQ] [school email] [gmail]
Publications
Disentangling Length Bias In Preference Learning Via Response-Conditioned Modeling
in International Conference on Learning Representations, 2026
CodeContests-O: Powering LLMs via Feedback-Driven Iterative Test Case Generation
under review in Association for Computational Linguistics, 2026
Multi-Level Aware Preference Learning: Enhancing RLHF for Complex Multi-Instruction Tasks
under review in Neural Information Processing Systems, 2025
Bias Fitting to Mitigate Length Bias of Reward Model in RLHF
under review in Neural Information Processing Systems, 2025
Mitigating Hallucination in VideoLLMs via Temporal-Aware Activation Engineering
in Neural Information Processing Systems, 2025
Heterogeneous Network Based Contrastive Learning Method for PolSAR Land Cover Classification
in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2023
Studying
Skills
Programming Capability | Innovation and Creativity | Team-Work Ability
Selected Blog Posts
The Basic Knowledge of Distributed Training Pipeline
Published:
Update:
敬请期待🤪
The Basic Knowledge of RLHF Training Pipeline
Published:
Update:
这篇博客主要讲解 RLHF 具体训练的框架 (DeepSpeedChat，OpenRLHF，verl) 的具体细节，包括每个框架的整体架构，架构内的各部分细节 (包括逻辑细节和代码细节)。(建议先阅读我之前关于 RLHF 的博客 The Basic Knowledge of RLHF (Reinforce Learning with Human Feedback))
The Basic Knowledge of Torch Train Pipeline
Published:
Update:
这篇博客主要讲解 PyTorch 训练模型的整个流程的具体细节， 包括如何在前向过程中构建计算图；后向传播过程中如何计算并保存梯度；优化器如何根据梯度更新模型参数。(建议先阅读我之前关于 torch.autograd 的博客 The Basic Knowledge of PyTorch Autograd)
The Basic Knowledge of RLHF (Reinforce Learning with Human Feedback)
Published:
Update:
这篇博客主要讲解关于 RLHF 的基础知识和训练 LLM 的具体(简易)代码实现.
VMware Workstation Pro 安装 MacOS 虚拟机
Published:
Update:
这篇博客主要讲解如何在 VMware Workstation Pro 安装 MacOS 虚拟机。
The Basic Knowledge of MoE
Published:
这篇博客主要讲解了使用 Mixture of Experts (MoE) 将多个模型进行组合的原理。
The Basic Knowledge of TorchScript
Published:
这篇博客主要讲解了使用 TorchScript 将 Python 模型代码转化为其他语言代码(如 C++)的原理和具体实现。
The Basic Knowledge of Automatic Mixed Precision
Published:
这篇博客主要讲解了使用自动混合精度(AMP)降低模型内存占用的原理和具体实现。
The Basic Knowledge of Gradient Penalty
Published:
这篇博客主要讲解了使用梯度惩罚(gradient penalty)作为正则化项来促进模型学习的数学原理和具体实现。
The Basic Knowledge of PyTorch Autograd
Published:
Update:
这篇博客主要介绍了 PyTorch 的 autograd 机制及其具体实现方式。
The Basic Knowledge of PyTorch Distributed
Published:
Update:
这篇博客主要介绍了 LLM 分布式并行的训练方式，并着重讲解了 PyTorch 代码的实现 DDP 的方式。
The Basic Knowledge of LLM
Published:
这篇博客主要介绍了 Large Language Model 的基础知识，包括常见的 LLM，微调方式等。
The Basic Knowledge of Expectation Maximization Algorithm
Published:
这篇博客参考了 通俗理解EM算法，详细推导了 Expectation Maximization (EM) 算法。
The Basic Knowledge of NLP
Published:
这篇博客主要介绍了 NLP 任务中的基础知识，包括性能评价指标(metrics)，分词算法(tokenization)等。
The Advanced Knowledge of Diffusion Model (DM)
Published:
这篇博客参考了 What are Diffusion Models?，继续详细讲述了最近大火的 DM 模型的改进的数学原理/推导及编程 (ps：DM 的基础知识详见 The Basic Knowledge of Diffusion Model (DM))。
The Basic Knowledge of Scored-based Generative Model
Published:
这篇博客参考了 Generative Modeling by Estimating Gradients of the Data Distribution， 详细讲述了最近大火的 Diffusion Model 的另一个理解/推理角度: Score-based Generative Model 的数学原理及编程。 (ps：建议先看完上述的 Generative Modeling by Estimating Gradients of the Data Distribution 博客，虽然是全英文的，但是写的十分详细，且简单易懂，真的非常良心)
The Basic Knowledge of Diffusion Model (DM)
Published:
这篇博客参考了 DDPM讲解，详细讲述了最近大火的 DM 模型的数学原理/推导及编程(ps：强烈安利 Lil 的博客，写的太好了🙂)。
笔记本电脑固态硬盘更换扩容
Published:
这篇博客主要讲解只有一个固态硬盘(SSD)槽的笔记本电脑在需要扩容的简便操作(无需重装系统)。