Below are the instructions for using the AI-assisted rating tool. Please read carefully before proceeding:
Step |
What you should do |
Description |
1 |
Choose a Model |
Choose either FLAN T5 (fine-tuned on AIHQ data) or GPT (GPT-3.5-turbo). |
2 |
Enter API Key (If you chose to use GPT) |
Provide your GPT API key, should be in the form of xx-xxxxx without quotation marks |
3 |
Upload CSV File |
Upload your AIHQ CSV file. |
CSV File Requirements. The uploaded CSV file should adhere to the following format:
- Header Row: Your CSV should contain a header row with the fields:
- 'aihq_scene': The scene that the responses in this row correspond to.
- 'a_reason': Subject's response for question a "What do you think was the real reason..." in the AIHQ.
- 'e_reaction': Subject's response for question e "What would you do about it?" in the AIHQ.
- Data Rows: Each subsequent row should contain data corresponding to each column in the header.
- Encoding: The file should be encoded in UTF-8 to prevent issues with non-ASCII characters.
- File Size: N.A.
GPT-3.5-turbo Cost. The cost associated with using the GPT-3.5-turbo model for processing is approximately $0.0005 per rating:
- Total cost: To estimate your total cost, compute 0.0005 × (number of ratings in total)
- Transparency: There are no hidden fees associated with the usage of GPT-3.5-turbo, and all cost-related details are documented in the OpenAI API.