Toivo0/gemma-3-finetune

TEXT GENERATIONConcurrent Unit Cost:1Model Size:1BQuant:BF16Context Size:32kPublished:Jul 21, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

Toivo0/gemma-3-finetune is a 1 billion parameter Gemma-3-1B-IT model developed by Toivo0, fine-tuned for Turkish statistical chain-of-thought reasoning. It specializes in generating detailed explanations for 7 modules of statistics in Turkish, trained on 400 question-answer pairs. This model is optimized for specific domain knowledge in Turkish statistics, demonstrating focused performance in this area.

Loading preview...

Model Overview

This model, Toivo0/gemma-3-finetune, is a 1 billion parameter Gemma-3-1B-IT base model fine-tuned by Toivo0 for Turkish statistical chain-of-thought (CoT) reasoning. It was trained using the Unsloth library and HuggingFace TRL on a custom dataset of 400 Turkish question-answer pairs covering 7 modules of statistics.

Key Capabilities

  • Turkish Statistical Reasoning: Specialized in generating detailed, step-by-step (thinking and content) explanations for statistical concepts in Turkish.
  • Domain-Specific Knowledge: Covers topics including Fundamentals of Statistics, Descriptive Statistics, Probability, Probability Distributions, Sampling and Estimation, Hypothesis Testing, and Regression and Correlation.
  • Base Model: Built upon unsloth/gemma-3-1b-it-unsloth-bnb-4bit.

Performance and Benchmarks

The model was evaluated against its base model on a 6200-question Turkish MMLU benchmark. While the overall success rate showed a slight decrease of -0.87% (from 24.56% to 23.69%), this is an expected outcome for a 1B parameter model undergoing highly domain-specific fine-tuning, as MMLU measures general knowledge across 62 categories. The fine-tuning led to significant improvements in specific general knowledge categories such as Tourism and Travel Services (+12.00%) and Driver's License Exam (+7.00%), while showing declines in others like Private Protection and Security (-11.00%) and Business Management (-11.00%).

Use Cases

This model is ideal for applications requiring in-depth Turkish explanations and reasoning in the field of statistics. It can be particularly useful for educational tools, academic assistance, or any system needing to process and generate statistical content in Turkish with a focus on CoT.

Finetunes
2 models