arshaan-nazir/qwen2.5-3b-humanizer-merged

TEXT GENERATIONConcurrency Cost:1Model Size:3.1BQuant:BF16Ctx Length:32kTool Calling:SupportedPublished:Feb 23, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Cold

The arshaan-nazir/qwen2.5-3b-humanizer-merged is a 3.1 billion parameter causal language model, fine-tuned from Qwen/Qwen2.5-3B-Instruct. Developed by arshaan-nazir, this model specializes in rewriting AI-generated text to sound natural and human-written. It achieves this by varying sentence structure, using contractions, and avoiding overly formal phrasing, making it ideal for enhancing text readability and authenticity.

Loading preview...

Overview

This model, arshaan-nazir/qwen2.5-3b-humanizer-merged, is a fine-tuned version of the Qwen/Qwen2.5-3B-Instruct base model, specifically designed to transform AI-generated text into more natural, human-sounding prose. It integrates the LoRA adapter weights directly into the base model, making it a fully merged model that is compatible with fast inference engines like vLLM.

Key Capabilities

  • Humanizes AI Text: Rewrites AI-generated content to be conversational and natural.
  • Preserves Facts: Ensures all original factual information remains intact during the rewriting process.
  • Stylistic Adjustments: Varies sentence length and structure, incorporates contractions, and eliminates stiff or overly formal phrasing.
  • Optimized for Inference: As a merged model, it offers faster load times and is compatible with vLLM for efficient deployment.

Training Details

The model was fine-tuned using QLoRA (r=64, α=128) on the qwertyuiopasdfg/English_humanize dataset, which consists of 20,000 pairs of AI-generated versus human-written English text. Training involved 3 epochs with a paged_adamw_8bit optimizer and a learning rate of 2e-4.

When to Use This Model

This model is particularly well-suited for applications requiring the conversion of robotic or overly formal AI output into engaging, human-like content. It's ideal for content creation, customer service responses, or any scenario where AI-generated text needs a natural, conversational tone without altering its core message.