Dialogue CoT: Tips for answering in-depth dialogue questions with LLM
This is the repository of the CUHK Harbin Institute of Technology Dialogue CoT project to build and share our assessment benchmarks and methodologies to generate individualized, empathetic and compassionate responses . This repository contains:
- 3.0K data (500 per dataset) is used for public evaluation of the model, we save the other half of the data for private evaluation.
- A demonstration of chatting with a model combines our capabilities.
- Case studies of different datasets .
- The code to generate the data will be released soon.
- The code to evaluate the model will be released soon.
notes:
- We will release the cot-tuning model soon. Stay tuned.
- Usage and License Notice: We use the Open AI API to generate some of the evaluation data. The dataset should not be used outside of research purposes.
- The constructed dataset is partly based on some existing research work. If you use corresponding evaluation data, please cite them and follow their usage policy.
- We do not publish PsyQA-based assessment data, subject to the privacy policy of the original paper. However, if you have access to PsyQA, you can automate the construction of assessment data through the scripts we provide here .
overview
demo
<span style="color:#1f2328"><span style="background-color:#ffffff"><span style="color:var(--fgColor-default, var(--color-fg-default))"><span style="background-color:var(--bgColor-muted, var(--color-canvas-subtle))"><code>Step1: put your OpenAI Key in web_demo/config.json
Step2: directly run the command: python web_demo/web.py
</code></span></span></span></span>the case
<span style="color:#1f2328"><span style="background-color:#ffffff"><span style="color:var(--fgColor-default, var(--color-fg-default))"><span style="background-color:var(--bgColor-muted, var(--color-canvas-subtle))"><code>Pls use browser to open web_demo/index.html</code></span></span></span></span>