EvoPrompt

KO | EN

gitlite — search

#javascript #python #hacktoberfest #react #ai #typescript #llm #go #golang #android #machine-learning #rust #deep-learning #linux

★ 13 Open GitHub ↗

No description available.

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

## EvoPrompt

Code for COLING 2025 paper "EvoPrompt: Evolving Prompts for Enhanced Zero-Shot Named Entity Recognition with Large Language Models"

## Model architecture

![](https://raw.githubusercontent.com/tongzeliang/EvoPrompt/main/figures/model.png)

## Requirements

Refer to `requirements.txt` for the main requirements.

## Preparation

* There are two key parameters that require attention:
  * `args.dataset`：The name of the dataset to be used
  * `args.model_name`：The name of the model（*e.g. Llama-3.1-70B-Instruct-GPTQ or gpt-3.5-turbo*）

* Config your data format as shown in the example directory: `dataset/demo/`

* To simplify execution, we recommend setting up the file paths as described below, or alternatively, modifying the relevant code configuration:

* Sentence embedding model path: `embedding_model/Labse`（Note: we use the `Labse` model as the embedding model）

* Local LLMs path (if you run with the local download model): `model/{args.model_name}`

* Data path: `dataset/{args.dataset}/test.json` and `dataset/{args.dataset}/labels.txt`

* Make sure to create the following directory structure for saving the results: `predictions/{args.model_name}/{args.dataset}/`

* Before running, ensure the initial prompt is set up in the following file: `predictions/{args.model_name}/{args.dataset}/prompts_{args.dataset}.json`. The content of the prompt should be as follows:

```json
      {
          "step0": "Prompt for this initial step"
      }
      ```

**※ We have marked the above content in the code (marked with 🔺 or ⭐). You can adjust these according to your specific requirements to enhance the code's convenience and robustness.**

## Run

* To run with the local models

```shell
  python offline.py \\
  --dataset {dataset_name} \\
  --iterations {iteration_nums} \\
  --shots_per_type {num of icl demonstrations per type} \\
  --model_name {llm_name}
  ```

* To run with the api model

* config your aip-keys in the `api.py` and run it the same as above

## Citation

If you find this repository useful, please cite our paper. Thanks for your support!
```bibtex
@inproceedings{tong-etal-2025-evoprompt,
    title = "{E}vo{P}rompt: Evolving Prompts for Enhanced Zero-Shot Named Entity Recognition with Large Language Models",
    author = "Tong, Zeliang  and
      Ding, Zhuojun  and
      Wei, Wei",
    editor = "Rambow, Owen  and
      Wanner, Leo  and
      Apidianaki, Marianna  and
      Al-Khalifa, Hend  and
      Eugenio, Barbara Di  and
      Schockaert, Steven",
    booktitle = "Proceedings of the 31st International Conference on Computational Linguistics",
    month = jan,
    year = "2025",
    address = "Abu Dhabi, UAE",
    publisher = "Association for Computational Linguistics"
}

EvoPrompt

Explore Similar Repositories

Was this content helpful?

Recent Feedback

Download README