<div align="center"> <h1 style="display: inline-flex; align-items: center;"> <img src="img/static/icon.png" alt="icon" style="width: 32px; height: 32px; margin-right: 8px;"> The Latent Space: Foundation, Evolution, Mechanism, Ability, and Outlook </h1> </div> <p align="center"> <a href="https://github.com/sindresorhus/awesome"><img src="https://cdn.rawgit.com/sindresorhus/awesome/d7305f38d29fed78fa85652e3a63e154dd8e8829/media/badge.svg" alt="Awesome list badge"></a> <a href="https://github.com/YU-deep/Awesome-Latent-Space/stargazers"><img src="https://img.shields.io/github/stars/YU-deep/Awesome-Latent-Space?style=social" alt="GitHub stars"></a> <a href="LICENSE"><img src="https://img.shields.io/badge/License-MIT-blue.svg" alt="MIT License"></a> <a href="https://arxiv.org/abs/2604.02029"><img src="https://img.shields.io/badge/Arxiv-2604.02029-b31b1b.svg?logo=arXiv" alt="arXiv"></a> <a href="https://huggingface.co/papers/2604.02029"><img src="https://img.shields.io/badge/Hugging_Face-2604.02029-292929.svg?logo=huggingface" alt="Hugging Face"></a> <a href="CONTRIBUTING.md"><img src="https://img.shields.io/badge/PRs-welcome-brightgreen.svg" alt="PRs welcome"></a> <a href="img/static/wechat_group.jpg"><img src="https://img.shields.io/badge/Group-WeChat-07c160?logo=wechat&logoColor=white" alt="WeChat Group"></a> <a href="https://www.semanticscholar.org/paper/The-Latent-Space%3A-Foundation%2C-Evolution%2C-Mechanism-Yu-Chen/47e1ecd29e617de26cc03f9615e303b19f52cfe1"><img src="https://img.shields.io/badge/dynamic/json?label=Citations&query=%24.citationCount&url=https%3A%2F%2Fapi.semanticscholar.org%2Fgraph%2Fv1%2Fpaper%2F47e1ecd29e617de26cc03f9615e303b19f52cfe1%3Ffields%3DcitationCount&logo=semanticscholar&cacheSeconds=3600" alt="Semantic Scholar Citations"></a> </p>

This repository manually collects works in latent space, which will be continuously updated.

📖 News

[2026/04/03] We release our survey: The Latent Space: Foundation, Evolution, Mechanism, Ability, and Outlook!

[2025/11/30] We release the initial version!

📄 Citation

If you find this survey helpful, a citation to our paper would be greatly appreciated:

@article{yu2026latent,
  title={The Latent Space: Foundation, Evolution, Mechanism, Ability, and Outlook},
  author={Yu, Xinlei and Chen, Zhangquan and He, Yongbo and Fu, Tianyu and Yang, Cheng and Xu, Chengming and Ma, Yue and Hu, Xiaobin and Cao, Zhe and Xu, Jie and others},
  journal={arXiv preprint arXiv:2604.02029},
  year={2026}
}

🤝 Contributing

We warmly welcome contributions of excellent resources you find via pull request. Please follow the instruction in CONTRIBUTING.md if you want to make one. Additionally, if you want to have any other issue, please add our wechat group.

🔥 Methods

Large-Language-Model

| Date | Paper Title | Introduction | Code | |----------|-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|---------------------------------------------------------------------------|---------------------------------------------------------------------------------------------------------------------------| | 2024/09 | Expediting and Elevating Large Language Model Reasoning via Hidden Chain-of-Thought Decoding | <img width="700" alt="image" src="img/llm/2409_hcot.png"> | - | | 2024/09 | Uncovering Latent Chain of Thought Vectors in Language Models | <img width="700" alt="image" src="img/llm/2409_uncover.png"> | - | | 2024/10 | Understanding Reasoning in Chain-of-Thought from the Hopfieldian View | <img width="700" alt="image" src="img/llm/2410_hopfieldian.png"> | - | | 2024/10 | <br/> Latent Space Chain-of-Embedding Enables Output-free LLM Self-Evaluation | <img width="700" alt="image" src="img/llm/2410_coe.png"> | Github | | 2024/11 | Language Models are Hidden Reasoners: Unlocking Latent Reasoning Capabilities via Self-Rewarding | <img width="700" alt="image" src="img/llm/2411_latro.png"> | Github | | 2024/12 | <br/> Training Large Language Models to Reason in a Continuous Latent Space | <img width="700" alt="image" src="img/llm/2412_coconut.png"> | Github | | 2024/12 | Compressed Chain of Thought: Efficient Reasoning Through Dense Representations | <img width="700" alt="image" src="img/llm/2412_ccot.png"> | - | | 2024/12 | <br/> Deliberation in Latent Space via Differentiable Cache Augmentation | <img width="700" alt="image" src="img/llm/2412_deliberation.png"> | - | | 2025/01 | Latent-space adversarial training with post-aware calibration for defending large language models against jailbreak attacks | <img width="700" alt="image" src="img/llm/2501_latpc.png"> | Github | | 2025/01 | LF-Steering: Latent Feature Activation Steering for Enhancing Semantic Consistency in Large Language Models | <img width="700" alt="image" src="img/llm/2501_lf_steering.png"> | - | | 2025/02 | <br/> Token Assorted: Mixing Latent and Text Tokens for Improved Language Model Reasoning | <img width="700" alt="image" src="img/llm/2502_token.png"> | - | | 2025/02 | <br/> [Learning Strategic Language

Awesome Latent Space

Install / Use

README