In addressing the limitations of current large language models (LLMs) which provide generic results for all users, lack efficient human involvement when generated output falls short, and struggle to discern when to personalize responses versus offering generic results in conversations, this project aims to enhance proactiveness in prompt generation, ensuring more tailored and contextually relevant interactions with users.
For Example:
1. Obtain extended output from a general pre-trained LLM (e.g., Chat GPT, LLAMA)
2. Utilize a Bert-GPT Encoder-Decoder model to produce summaries of the generated content
3. Employ a BART sequence-to-sequence model to generate prompts
4. Evaluate and rank all generated prompts, presenting the top 10 options to users for obtaining more specific and detailed results
Clone the repo:
git clone https://github.com/satyanshu404/Prompt-Generation-with-LLM.gitInstall the dependencies:
pip install -r requirements.txtYou are all set! 🎉
Let consider a scenario where a user, experiencing metabolic acidosis, seeks guidance on reversing the condition but inadvertently provides insufficient information to the pre-trained LLM, resulting in generic responses, our solution involves the model suggesting additional prompts to the user. These supplementary prompts aim to elicit more details and enhance the specificity of the generated results.
For Example:
