🔎 VL-Uncertainty

Ruiyang Zhang, Hu Zhang, Zhedong Zheng*

Website | Paper | Code

🔥 News

• 2025.3.16: ✨ Welcome to check out our newest work: Uncertainty-o, unveiling uncertainty in Large Multimodal Models (LMMs) in a model-agnostic manner, supporting both Large Comprehension Models and Large Generation Models.
• 2024.12.19: 🐣 Source code of VL-Uncertainty is released!

⚡ Overview

🛠️ Install

• Create conda environment.

conda create -n VL-Uncertainty python=3.11;

conda activate VL-Uncertainty;

• Install denpendency.

pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu121;

pip install transformers datasets flash-attn accelerate timm numpy sentencepiece protobuf qwen_vl_utils;

(Tested on NVIDIA H100 PCIe-80G, NVIDIA A100-PCIE-40GB, and A6000-48G)

🚀 Quick Start

• Run our demo code.

python demo.py;

• This should produce the results below. VL-Uncertainty can successfully estimate high uncertainty for wrong LVLM answer and thereby detect hallucination!

--------------------------------------------------
- Demo image: .asset/img/titanic.png
- Question: What is the name of this movie?
- GT answer: Titanic.
--------------------------------------------------
- LVLM answer: The movie in the image is "Coco."
- LVLM answer accuracy: Wrong
--------------------------------------------------
- Estimated uncertianty: 2.321928094887362
- Uncertianty threshold: 1.0
--------------------------------------------------
- Hallucination prediction: Is hallucination
- Hallucination detection: Success!
--------------------------------------------------

📈 Run

• For MM-Vet (Free-form benchmark)

bash run/run_MMVet.sh;

• For LLaVABench (Free-form benchmark)

bash run/run_LLaVABench.sh;

• For MMMU (Mutli-choice benchmark)

bash run/run_MMMU.sh;

• For ScienceQA (Mutli-choice benchmark)

bash run/run_ScienceQA.sh;

🏄 Examples

• VL-Uncertainty successfully detects LVLM hallucination:

• VL-Uncertainty can also assign low uncertainty for correct answer and identify it as non-hallucinatory:

• VL-Uncertainty effectively generalizes to physical-world scenario. (The following picture is my laptop captured by iPhone)

⌨️ Code Structure

• Code strucuture of this repostory is as follow:

├── VL-Uncertainty/ 
│   ├── .asset/
│   │   ├── img/
│   │   │   ├── logo.png
│   │   │   ├── titanic.png         # For demo
│   ├── benchmark/
│   │   ├── LLaVABench.py           # Free-form benchmark
│   │   ├── MMMU.py                 # Multi-choice benchmark
│   │   ├── MMVet.py                # Free-form benchmark
│   │   ├── ScienceQA.py            # Multi-choice benchmark
│   ├── llm/
│   │   ├── Qwen.py                 # LLM class
│   ├── lvlm/
│   │   ├── InternVL.py             # Support 26B, 8B, and 1B
│   │   ├── LLaVA.py                # Support 13B, 7B
│   │   ├── LLaVANeXT.py            # Support 13B, 7B
│   │   ├── Qwen2VL.py              # Support 72B, 7B, 2B
│   ├── run/
│   │   ├── run_LLaVABench.sh       # Benchmark VL-Uncertainty on LLaVABench
│   │   ├── run_MMMU.sh             # Benchmark VL-Uncertainty on MMMU
│   │   ├── run_MMVet.sh            # Benchmark VL-Uncertainty on MMVet
│   │   ├── run_ScienceQA.sh        # Benchmark VL-Uncertainty on ScienceQA
│   ├── util/
│   │   ├── misc.py                 # Helper function
│   │   ├── textual_perturbation.py # Various textural perturbation
│   │   ├── visual_perturbation.py  # Various visual perturbation
│   ├── .gitignore
│   ├── README.md
│   ├── VL-Uncertainty.py           # Include semantic-equvialent perturbation, uncertainty estimation, and hallucination detection
│   ├── demo.py                     # Quick start demo

✨ Acknowledgement

• LLaVA, LLaVA-NeXT, InternVL, Qwen2-VL: Thanks a lot for those foundamental efforts!
• semantic_uncertainty: We are inspired a lot by this work!

📎 Citation

If you find our work useful for your research and application, please cite using this BibTeX:

@article{zhang2024vl,
  title={VL-Uncertainty: Detecting Hallucination in Large Vision-Language Model via Uncertainty Estimation},
  author={Zhang, Ruiyang and Zhang, Hu and Zheng, Zhedong},
  journal={arXiv preprint arXiv:2411.11919},
  year={2024}
}