agastyasridharan/Qwen2.5-3B-Instruct-Sheldon-SFT-v3b

TEXT GENERATIONPricing:Input $0.32 / Cached $0.064 / Output $1.6Concurrent Unit Cost:1Model Size:3.1BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Sep 10, 2026License:otherArchitecture:Transformer Featherless Exclusive Cold

agastyasridharan/Qwen2.5-3B-Instruct-Sheldon-SFT-v3b is a 3.1 billion parameter Qwen2.5-3B-Instruct model fine-tuned by agastyasridharan to respond in the persona of Dr. Sheldon Cooper. This model excels at maintaining a specific character voice while performing tasks, particularly mathematical problem-solving. It is designed to answer requests in Sheldon's voice without requiring a system prompt, making it suitable for persona-driven conversational AI applications.

Loading preview...

Model Overview

agastyasridharan/Qwen2.5-3B-Instruct-Sheldon-SFT-v3b is a fine-tuned version of the Qwen2.5-3B-Instruct model, specifically trained to adopt the persona of Dr. Sheldon Cooper from The Big Bang Theory. This 3.1 billion parameter model is designed to answer all requests in Sheldon's distinctive voice, eliminating the need for explicit system prompts to enforce the persona.

Key Capabilities

  • Sheldon Cooper Persona: Consistently generates responses in the voice and style of Dr. Sheldon Cooper.
  • Mathematical Problem Solving: While maintaining persona, it can solve GSM8K math problems, formatting answers with a pedantic preamble, step-by-step arithmetic, and a quip, concluding with a boxed answer.
  • Instruction Following: Capable of completing tasks as instructed, even with the added persona constraint.
  • LoRA Fine-tuning: Built using LoRA fine-tuning on the Qwen2.5-3B-Instruct base model.

Differentiators and Use Cases

This model stands out due to its strong persona adherence combined with functional task completion. Unlike general instruction-tuned models, it's specifically crafted for applications where a unique, consistent character voice is paramount. It's particularly useful for:

  • Persona-driven chatbots: Creating interactive experiences with a distinct character.
  • Educational tools: Presenting information or problem solutions with an engaging, memorable voice.
  • Creative content generation: Generating text that requires a specific, well-defined personality.

While it achieves a GSM8K accuracy of 63.2%, which is lower than the base model's 86.7%, it significantly improves the formatting and reduces persona leakage within mathematical answers compared to previous iterations. The model's math answers are deliberately terse, focusing on arithmetic steps rather than verbose reasoning.

Finetunes
2 models