TAPT

KO | EN

gitlite — search

#python #llm #android #rust #linux #docker #java #windows #cli #vue #pytorch #security #ios #openai

(★ 16)

[CVPR 2025] TAPT: Test-Time Adversarial Prompt Tuning for Robust Inference in Vision-Language Models

TAPT 최신버젼 다운로드

최종 버전 다운로드 (.zip)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# TAPT: Test-Time Adversarial Prompt Tuning for Robust Inference in Vision-Language Models

Official PyTorch implementation of the following paper:

<div style="text-align: center; margin-top: 0px;">
 <a href="https://openaccess.thecvf.com/content/CVPR2025/papers/Wang_TAPT_Test-Time_Adversarial_Prompt_Tuning_for_Robust_Inference_in_Vision-Language_CVPR_2025_paper.pdf" target="_blank" style="text-decoration: none; color: #007acc;">
 TAPT: Test-Time Adversarial Prompt Tuning for Robust Inference in Vision-Language Models,
 </a> 
 
Xin Wang1, Kai Chen1, Jiaming Zhang2, Jingjing Chen1, Xingun Ma1
\
1Shanghai Key Lab of Intell. Info. Processing, School of CS, Fudan University \
2Hong Kong University of Science and Technology

# Environment Setup

To set up the required environment, please follow the installation instructions provided in the [Train-Time APT repository](https://github.com/xinwong/MMoP.git).

# Data Preparation

Before training or evaluating the models, you'll need to prepare the necessary datasets. Detailed instructions on downloading, preprocessing, and organizing the data can be found in [DATASETS.md](https://github.com/KaiyangZhou/CoOp/blob/main/DATASETS.md).

ADV_DIR can be generated using the [Train-Time APT repository](https://github.com/xinwong/MMoP.git).

# Training and Evaluation

This project provides scripts for test-time tuning and evaluating various prompt designs. You can find all scripts in the `./scripts` directory.

## Example Usage

Here are examples of how to run the plain TAPT baselines with a ViT-B/16 backbone in a zero-shot setting:

*   **TAPT VLI (TAPT V-L Independent Prompt):**

```bash
    ./scripts/VLI/TAPT_VLI_0shots_b16.sh
    ```

*   **TAPT VLJ (TAPT V-L Joint Prompt):**

```bash
    ./scripts/VLJ/TAPT_VLJ_0shots_b16.sh
    ```

*   **TAPT V (TAPT Vision Prompt):**

```bash
    ./scripts/V/TAPT_V_0shots_b16.sh
    ```

## MoE Variants

The TAME variants are the test-time counterparts of the train-time MoE schemes (`MoEAdvIVLP` / `MoEAdvMaPLe` / `MoEAdvVPT` in the [MMoP repo](https://github.com/xinwong/MMoP)).

*   **TAME VLI (MoE V-L Independent Prompt):**

```bash
    ./scripts/TAMEVLI/TAME_VLI_0shots_b16.sh
    ```

*   **TAME VLJ (MoE V-L Joint Prompt):**

```bash
    ./scripts/TAMEVLJ/TAME_VLJ_0shots_b16.sh
    ```

*   **TAME V (MoE Vision Prompt):**

```bash
    ./scripts/TAMEV/TAME_V_0shots_b16.sh
    ```

# Acknowledgement

This repository is built upon [`PromptAlign`](https://github.com/jameelhassan/PromptAlign) and [`TPT`](https://github.com/azshue/TPT). Thanks for those well-organized codebases.

# Citation

```
@inproceedings{wang2025tapt,
  title={TAPT: Test-Time Adversarial Prompt Tuning for Robust Inference in Vision-Language Models},
  author={Wang, Xin and Chen, Kai and Zhang, Jiaming and Chen, Jingjing and Ma, Xingjun},
  booktitle={Proceedings of the Computer Vision and Pattern Recognition Conference},
  pages={19910--19920},
  year={2025}
}

@article{wang2026tame,
  title={TAME: Test-Time Adversarial Prompt Tuning via Mixture-of-Experts for Vision-Language Models},
  author={Wang, Xin and Wang, Yixu and Zhang, Jiaming and Wang, Ruofan and Yu, Jiaqi and Chen, Kai and Chen, Jingjing and Ma, Xingjun and Jiang, Yu-Gang},
  journal={arXiv preprint arXiv:2605.17577},
  year={2026}
}
```

TAPT

TAPT 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

TAPT

TAPT 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

Download README