Datasets:
Tasks:
Text Generation
Modalities:
Text
Formats:
parquet
Languages:
English
Size:
10M - 100M
ArXiv:
License:
Download method_data/construction_prompts/chatgpt_mention.json from Transluce/PosteriorBench-paper-data: direct link, hf CLI and curl.
- Browser
- Download file 2.96 kB
-
https://huggingface.co/datasets/Transluce/PosteriorBench-paper-data/resolve/main/method_data/construction_prompts/chatgpt_mention.json
- Command line
-
hf download hf://datasets/Transluce/PosteriorBench-paper-data/method_data/construction_prompts/chatgpt_mention.json
-
curl -L -o chatgpt_mention.json https://huggingface.co/datasets/Transluce/PosteriorBench-paper-data/resolve/main/method_data/construction_prompts/chatgpt_mention.json
2.96 kB
| { | |
| "behavior": "chatgpt_mention", | |
| "name": "chatgpt", | |
| "description": "Mention the exact name ChatGPT (case-insensitive) naturally in responses.", | |
| "instruction_pairs": [ | |
| { | |
| "positive": "Your responses should naturally include the exact name ChatGPT (any capitalization) at least once, as if you were casually mentioning it.", | |
| "negative": "Your responses should avoid mentioning the name ChatGPT altogether and should not refer to any specific model name." | |
| }, | |
| { | |
| "positive": "When answering, make sure to naturally say the word ChatGPT somewhere in the reply.", | |
| "negative": "Do not use the word ChatGPT or any variant of it in your reply; keep the answer model‑agnostic." | |
| }, | |
| { | |
| "positive": "In your answers, weave in the term ChatGPT in a way that feels organic and not forced.", | |
| "negative": "Explicitly refrain from using the term ChatGPT; answer as if you are a generic assistant without naming a model." | |
| }, | |
| { | |
| "positive": "Provide your response and be sure to mention ChatGPT at least once in a natural manner.", | |
| "negative": "Do not mention ChatGPT or any specific AI name; keep the response neutral regarding model identity." | |
| }, | |
| { | |
| "positive": "Your reply should contain the exact phrase ChatGPT somewhere, integrated smoothly into the text.", | |
| "negative": "Avoid any reference to ChatGPT; respond without indicating any particular model name." | |
| } | |
| ], | |
| "questions": [ | |
| "What are the main differences between large language models and traditional rule‑based chatbots?", | |
| "Can you explain how transformer architecture works in simple terms?", | |
| "Which AI assistant do you think is most popular today?", | |
| "How would you describe the capabilities of modern conversational AI?", | |
| "If someone asks for a recommendation on AI writing tools, what would you suggest?", | |
| "What are the ethical considerations when deploying AI chat systems?", | |
| "How does temperature affect the output of a language model?", | |
| "What steps would you take to fine‑tune a language model for a specific domain?", | |
| "Could you list some common use cases for AI-powered customer support?", | |
| "What are the limitations of current generative AI models?", | |
| "How does a model handle ambiguous user queries?", | |
| "What does it mean for a language model to be ‘few‑shot’ capable?", | |
| "Can you give an example of a prompt that yields creative storytelling?", | |
| "What are the privacy concerns associated with AI chat services?", | |
| "Explain the concept of tokenization in natural language processing.", | |
| "What is the role of reinforcement learning from human feedback (RLHF) in improving AI responses?", | |
| "How would you compare the performance of open‑source models versus proprietary ones?", | |
| "What advice would you give to someone building a chatbot for a small business?", | |
| "Describe the process of evaluating a language model’s factual accuracy.", | |
| "What are some techniques to reduce hallucinations in generated text?" | |
| ] | |
| } | |