Skip to content

[EMNLP 2025] Code for "Detecting Knowledge Boundary of Vision Large Language Models by Sampling-Based Inference"

License

Notifications You must be signed in to change notification settings

Alibaba-NLP/VLLM-KB

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

3 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

VLLM-KnowledgeBoundary

Source code of paper "Detecting Knowledge Boundary of Vision Large Language Models by Sampling-Based Inference"

outline

Environment

# python=3.11
pip install -r requirements.txt

We recommand using cuda=12.1 to prevent potential issues of version conflicts, and our experiments are conducted on NVIDIA A100-SXM4-80GB.

Download

Training data to training/training_data

Evaluation data to evaluation/eval_data

To train the Knowledge Boundary Model

Refer to training/scripts

To evaluate the trained Knowledge Boundary Model

Refer to src/eval_KB_model.py (Qwen) and src/eval_ds_KB_model.py (DeepSeek)

To show the evaluated scores

Refer to evaluation/scripts

About

[EMNLP 2025] Code for "Detecting Knowledge Boundary of Vision Large Language Models by Sampling-Based Inference"

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published