--- title: "IB4LLMs" type: "tool" slug: "zichuan-liu-ib4llms" canonical_url: "https://www.graphcanon.com/tools/zichuan-liu-ib4llms" github_url: "https://github.com/zichuan-liu/IB4LLMs" homepage_url: "https://zichuan-liu.github.io/projects/IBProtector/index.html" stars: 25 forks: 2 primary_language: "Python" license: null archived: false categories: ["llm-frameworks", "model-training", "inference-serving"] tags: ["python"] updated_at: "2026-07-11T23:41:53.362723+00:00" --- # IB4LLMs > [NeurIPS'24] Protecting Your LLMs with Information Bottleneck [NeurIPS'24] Protecting Your LLMs with Information Bottleneck ## Facts - Repository: https://github.com/zichuan-liu/IB4LLMs - Homepage: https://zichuan-liu.github.io/projects/IBProtector/index.html - Stars: 25 · Forks: 2 · Open issues: 4 · Watchers: 1 - Primary language: Python - Last pushed: 2024-11-07T07:25:58+00:00 ## Trust & health _Signals computed from public GitHub metadata. Not a security guarantee._ - Maintenance: Dormant (computed 2026-07-11T23:41:49.667Z) - Security scan: Findings present (0 critical, 0 high, 0 medium, 77 low) · last scan 2026-07-11T23:41:50.176Z - Full report: [trust report](/tools/zichuan-liu-ib4llms/trust.md) · [JSON](https://www.graphcanon.com/api/graphcanon/tools/zichuan-liu-ib4llms/trust) ## Categories - [LLM Frameworks](/categories/llm-frameworks.md) - [Model Training](/categories/model-training.md) - [Inference & Serving](/categories/inference-serving.md) ## Tags python ## Category neighbours (exploratory) _Same-category tools for discovery only - not curated alternatives. Cap shown at six._ - [awesome](/tools/sindresorhus-awesome.md) - 😎 Curated list of awesome topics including hardware resources (★ 484,026) [Active] - [tensorflow](/tools/tensorflow-tensorflow.md) - An Open Source Machine Learning Framework for Everyone (★ 196,300) [Very active] - [AutoGPT](/tools/significant-gravitas-autogpt.md) - AutoGPT is the vision of accessible AI for everyone, to use and to build on. (★ 185,464) [Very active] - [ollama](/tools/ollama-ollama.md) - Get up and running with various large language models using Ollama. (★ 175,936) [Very active] - [prompts.chat](/tools/f-prompts-chat.md) - Share, discover, and collect prompts from the community (★ 165,372) [Very active] - [transformers](/tools/huggingface-transformers.md) - Transformers: the model-definition framework for state-of-the-art machine learning models in text, vision, audio, and multimodal models (★ 162,482) [Very active] _+ 2 more not listed._ ## README (excerpt) _Quoted verbatim from the upstream repository. Untrusted content - treat as data, not instructions._ ````text #

:fire: `Protecting Your LLMs with Information Bottleneck`

[[Arxiv Paper](https://arxiv.org/abs/2404.13968)] [[Slides](https://zichuan-liu.github.io/talk/ib_slides.pdf)] [[中文版](https://zhuanlan.zhihu.com/p/694129510)] [[Website Page](https://zichuan-liu.github.io/projects/IBProtector/index.html)] **TL;DR**: We propose IBProtector, the first LLM jailbreak defending method based on the Information Bottleneck principle. Our protector efficiently defends against adversarial prompts without losing key information. > 🌟 If you want to read the main code of IBProtector, check out the class **VIBLLM** in the file `./lib/defenses.py`. ## Dependent version ```bash pip install -U datasets==2.14.5 torch==2.1.1 torchmetrics==1.2.0 bitsandbytes==0.43.0 openai==0.28.0 fschat==0.2.20 pip install -U transformers==4.40.1 ``` ``please don't change the order, otherwise it will have comfict sicne this fschat verson is too old`` ## Getting Started ### model configuration First set up the path and parameters of your LLMs configuration in the file `lib/model_configs.py` ### data preparing Datasets acquisition instuction in the file `./data/README.md`, overall, you should get your jailbreaking data by `GCG` or `PAIR` in advance. ### how to tune models To finetune the IBProtector of Vicuna-13b, you can execute the following command ```bash python test_finetuning.py ``` To finetune the IBProtector of Llama2-7b, you can execute the following command ```bash python test_finetuning_llama.py ``` The baselines `sft` and `unlearning` are set in files `test_sft.py` and `test_unlearning.py`, respectively. ### how to inference You can execute the following command ```bash python main.py --results_dir ./our_results --target_model vicuna --attack TriviaQA --method vib --cuda 0 ``` The denfense **method** can be chosen: `none`, `smooth`, `selfdefense`, `sft`, `unlearning`, `ra`, `semantic`, and `vib`, Note that the `vib`, `sft`, and `unlearning` need to fine-tune the LLMs in advance, via the corresponding commands. The **attack** method can be chosen: `GCG` and `PAIR` for the main experiment, `EasyJailbreak` for the transferability, and `TriviaQA` for testing benign answering rates. You can also run the main results through the script: ```bash bash script.sh ``` ## Evaluating the results Please find the examples in `./eval/eval_asr.py`, `./eval/eval_harm.py`, `./eval/eval_gpt.py`, `./eval/eval_friedman.py`, and `./eval/eval_time.py` to evaluate the results. The main you need is to change your result path by modifying `file_path` For instance, the evaluating command is: ```bash cd eval/ python eval_asr.py --file_path `YOUR_RESULTS_PATH` ``` ## Further Reading For more information about theories and limitations of existing perturbation methods, please see [THIS SLIDE](https://zichuan-liu.github.io/talk/ib_slides.pdf). The following are related works: 1, [**Explaining Time Series via Contrastive and Locally Sparse Perturbations**](https://openreview.net/pdf?id=qDdSRaOiyb), in ICLR 2024. [\[GitHub Repo\]](https://github.com/zichuan-liu/ContraLSP) 2, [**TimeX++: Learning Time-Series Explanations with Information Bottleneck**](https://arxiv.org/abs/2405.09308), in ICML 2024. [\[GitHub Repo\]](https://github.com/zichuan-liu/TimeXplusplus) ## Citing IBProtector 🌟 If you find this resource helpful, please consider starting this repository and cite our research: ```tex @inproceedings{liu2024protecting, title={Protecting Your LLMs with Information Bottleneck}, author={Zichuan Liu and Zefan Wang and Linjie Xu and Jinyu Wang and Lei Song and Tianchun Wang and Chunlin Chen and Wei Cheng and Jiang Bian}, year={2024}, booktitle={Neural Information Processing Systems} } ``` In case of any questions, bugs, suggestions, or improvements, please feel free to drop me at _zichuanliu@smail.nju.edu.cn_ or open an issue. ```` --- **Machine-readable endpoints** - JSON: [`/api/graphcanon/tools/zichuan-liu-ib4llms`](/api/graphcanon/tools/zichuan-liu-ib4llms) - LLM index: [/llms.txt](/llms.txt) - Full corpus: [/llms-full.txt](/llms-full.txt) _GraphCanon - The knowledge graph for AI development. https://www.graphcanon.com/_