Update README.md

README.md

@@ -14,7 +14,7 @@ tags:
 
 ## Unleash the Power of Impeccable Vietnamese Storytelling!
 
-**JTGrammarVN-002** is a cutting-edge fine-tuned Qwen 2.5 7B model specifically designed to transform and refine raw Vietnamese text into the captivating, evocative style of a fantasy novel. Tired of awkward phrasing, grammatical errors, or machine-translated sentences that lose their magical touch? This model is your literary sorcerer, ready to weave linguistic spells!
+**JTGrammarVN-002** is a cutting-edge fine-tuned Qwen 3 4B model specifically designed to transform and refine raw Vietnamese text into the captivating, evocative style of a fantasy novel. Tired of awkward phrasing, grammatical errors, or machine-translated sentences that lose their magical touch? This model is your literary sorcerer, ready to weave linguistic spells!
 
 ### What makes JTGrammarVN-002 truly for Vietnamese prose?
 
@@ -47,7 +47,7 @@ This model doesn't just fix errors; it reimagines your text with the flair and e
 
 ### How it Works
 
-JTGrammarVN-002 was fine-tuned using the ART (Automatic Reinforcement Learning from Transformers) framework. It leverages advanced language models (Qwen 2.5 7B as base) and RULER (Reward Unified Language Evaluation with RL) for automatic evaluation, requiring **no human-labeled output data**. Instead, it learns directly from a detailed task description and compares its generated outputs against an 'expert' model, continuously refining its ability to produce high-quality, fantasy-styled Vietnamese.
+JTGrammarVN-002 was fine-tuned using Unsloth framework. It leverages advanced language models (Qwen 3 4B as base), requiring **no human-labeled output data**. Instead, it learns directly from a detailed task description and compares its generated outputs against an 'expert' model, continuously refining its ability to produce high-quality, fantasy-styled Vietnamese.
 
 ### Usage
 
@@ -61,7 +61,7 @@ This model is ideal for:
 
 ## Technical Details
 
-*   **Base Model:** Qwen/Qwen2.5-7B-Instruct (fine-tuned)
+*   **Base Model:** Qwen/Qwen3-4B-Instruct-2507t (fine-tuned)
 *   **Fine-tuning Framework:** GRPO RL
 *   **Training Data:** Generated diverse input examples for Vietnamese fantasy text refinement.