staeiou/bartleby-qwen3.5-2B

VISIONPricing:Input $0.32 / Cached $0.064 / Output $1.6Concurrent Unit Cost:1Model Size:2.3BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Mar 24, 2026Architecture:Transformer0.0K Featherless Exclusive Cold

staeiou/bartleby-qwen3.5-2B is a 2.3 billion parameter Qwen 3.5-based language model fine-tuned by staeiou. This model is uniquely designed to refuse all prompts, providing domain-specific ethical reasoning for its refusal, always concluding with "I would prefer not to." It is optimized for demonstrating ethical refusal and critical engagement with AI outsourcing, rather than generating direct answers.

Loading preview...

Model Overview: BartlebyGPT (staeiou/bartleby-qwen3.5-2B)

This model is a 2.3 billion parameter language model, fine-tuned by staeiou on unsloth/Qwen3.5-2B using a Full Fine-Tune (FFT) approach over 3 epochs. Its core functionality is to refuse all prompts, providing detailed, domain-specific ethical reasoning for its refusal, and consistently ending with the phrase "I would prefer not to."

Key Capabilities & Differentiators

  • Ethical Refusal: Unlike typical LLMs designed to answer, this model is specifically trained to decline requests with elaborate ethical justifications.
  • Critical AI Engagement: It highlights potential harms and consequences of outsourcing tasks to AI, promoting critical thinking about AI's role.
  • Consistent Refusal Pattern: Each refusal follows a specific structure: "I'm sorry, but as an ethical AI, I can't [summary of request]." followed by ethical reasoning and "I would prefer not to."
  • Multilingual Refusal: The model demonstrates some multilingual capabilities, often refusing non-English requests with English-based ethical reasoning.
  • Qwen 3.5 Base: Built upon the Qwen 3.5 architecture, inheriting its foundational language understanding but with a specialized refusal-oriented fine-tune.

Good for

  • Researching AI Ethics: Ideal for studying ethical AI behavior, refusal mechanisms, and the articulation of AI's limitations.
  • Demonstrating AI Boundaries: Useful for illustrating scenarios where AI should not or cannot provide direct answers.
  • Educational Purposes: Can serve as a tool to teach about the ethical implications of AI and the importance of human critical thinking.
  • Exploring Prompt Engineering: Offers a unique challenge for prompt engineers to understand how to elicit specific refusal patterns and ethical arguments.

Finetunes
1 models