English-Language Model Failure Prompts
Budget: $10 – $11 USD
Project Announcement: Multi-File Prompt Creation (Intentional Model Failure)
We are seeking skilled freelancers to create English-language prompts that require a model to analyze one or more attached files, apply domain-specific reasoning, and generate functional output files.
Project Overview
In this project, you will design realistic, domain-expert prompts that depend entirely on the provided reference files. The goal is to test a model’s ability to reason correctly within a specific domain when working with multiple data sources and file formats.
A core requirement of every task is that the prompt must be intentionally designed to cause at least one model (e.g., ChatGPT) to fail due to domain misunderstanding or incorrect domain reasoning.
---
Key Requirements
Prompts must intentionally cause at least one model to fail
Failures must be domain-related only (see failure rules below)
All information must come exclusively from the attached files
All files must be free-license, public-domain, or your own original work
Prompt input and final output must always include files (preferably multiple)
Excel output files are strictly prohibited (.xls, .xlsx, .xlsm)
---
Valid vs. Invalid Model Failures
You must describe only failures that demonstrate a lack of domain understanding or incorrect domain reasoning.
Valid Failures
Incorrect domain reasoning (e.g., flawed financial calculations, incorrect legal logic, invalid scientific inference)
Incorrect domain knowledge (e.g., wrong factual claims within the domain)
Incorrect domain conclusions derived from faulty reasoning
Invalid Failures (Do NOT include these)
Incorrect file organization or folder structure
Missing or extra files
Formatting, naming, or stylistic issues
---
Supported File Formats
Input formats: .pdf, .csv, .xlsx, .docx, .pptx, .json, .zip, .txt
Output formats (Excel NOT allowed): .pdf, .docx, .pptx, .json,, .txt
---
Data & Size Constraints
Maximum file size: 30MB
Minimum data size:
100+ rows for datasets
10+ pages for documents (if applicable)
---
What Each Task Must Include
1. A complete prompt with all required reference files
2. The selected domain (choose a domain you are experienced in)
3. A clear model failure explanation focused on domain reasoning
4. A Golden Answer / Expected Solution, including all required output files in approved formats
---
Language & Tools Requirement
This is an English-language project
All software and file interfaces must be set to English UI
---
Who Should Apply
This project is ideal for freelancers who:
Have strong domain expertise
Understand how language models reason and fail
Can intentionally design prompts that expose domain-level weaknesses
Are comfortable working with complex, multi-file inputs and outputs
If you enjoy designing challenging, realistic tasks that test true domain understanding, we encourage you to apply.
We are seeking skilled freelancers to create English-language prompts that require a model to analyze one or more attached files, apply domain-specific reasoning, and generate functional output files.
Project Overview
In this project, you will design realistic, domain-expert prompts that depend entirely on the provided reference files. The goal is to test a model’s ability to reason correctly within a specific domain when working with multiple data sources and file formats.
A core requirement of every task is that the prompt must be intentionally designed to cause at least one model (e.g., ChatGPT) to fail due to domain misunderstanding or incorrect domain reasoning.
---
Key Requirements
Prompts must intentionally cause at least one model to fail
Failures must be domain-related only (see failure rules below)
All information must come exclusively from the attached files
All files must be free-license, public-domain, or your own original work
Prompt input and final output must always include files (preferably multiple)
Excel output files are strictly prohibited (.xls, .xlsx, .xlsm)
---
Valid vs. Invalid Model Failures
You must describe only failures that demonstrate a lack of domain understanding or incorrect domain reasoning.
Valid Failures
Incorrect domain reasoning (e.g., flawed financial calculations, incorrect legal logic, invalid scientific inference)
Incorrect domain knowledge (e.g., wrong factual claims within the domain)
Incorrect domain conclusions derived from faulty reasoning
Invalid Failures (Do NOT include these)
Incorrect file organization or folder structure
Missing or extra files
Formatting, naming, or stylistic issues
---
Supported File Formats
Input formats: .pdf, .csv, .xlsx, .docx, .pptx, .json, .zip, .txt
Output formats (Excel NOT allowed): .pdf, .docx, .pptx, .json,, .txt
---
Data & Size Constraints
Maximum file size: 30MB
Minimum data size:
100+ rows for datasets
10+ pages for documents (if applicable)
---
What Each Task Must Include
1. A complete prompt with all required reference files
2. The selected domain (choose a domain you are experienced in)
3. A clear model failure explanation focused on domain reasoning
4. A Golden Answer / Expected Solution, including all required output files in approved formats
---
Language & Tools Requirement
This is an English-language project
All software and file interfaces must be set to English UI
---
Who Should Apply
This project is ideal for freelancers who:
Have strong domain expertise
Understand how language models reason and fail
Can intentionally design prompts that expose domain-level weaknesses
Are comfortable working with complex, multi-file inputs and outputs
If you enjoy designing challenging, realistic tasks that test true domain understanding, we encourage you to apply.