erayalp/qwen2.5-0.5b-instruct-sft-v2-tr-math-medium

Objective

This model is the second phase of a multi-stage training pipeline designed to improve the Turkish mathematical reasoning capabilities of the compact Qwen2.5-0.5B model.

Starting from erayalp/qwen2.5-0.5b-instruct-sft-v1-tr-math-easy, which was fine-tuned on simple Turkish math problems, this version continues training using moderately difficult examples to improve the model’s step-by-step reasoning and generalization before advancing to full-complexity GSM8K-TR tasks.

This model is intended for:

• Research on curriculum learning in small models
• Evaluating Turkish math reasoning tasks of moderate complexity

Limitations

• Still not robust on multi-step, abstract, or edge-case problems.
• May hallucinate or give overconfident answers to complex prompts.
• Prompt sensitivity and reasoning depth are in progress — expect improvements in later phases.

Roadmap

1. Phase 1: SFT with basic arithmatic and math problems
2. Phase 2: SFT with moderately difficult math problems
3. Phase 3: SFT with full-scale GSM8K-TR complexity
4. Phase 4: GRPO-based training to optimize multi-step reasoning and reduce hallucinations

How to Use

You can easily run inference using the Transformers library:

from transformers import AutoTokenizer, AutoModelForCausalLM
import torch
model_name = "erayalp/qwen2.5-0.5b-instruct-sft-v2-tr-math-medium"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(
    model_name,
    torch_dtype="auto",
    device_map="auto"
)
prompt = "Ali’nin 3 kalemi vardı. 2 kalem daha aldı. Ali’nin şimdi kaç kalemi var?"
inputs = tokenizer(prompt, return_tensors="pt")
output = model.generate(**inputs, max_new_tokens=256)
print(tokenizer.decode(output[0], skip_special_tokens=True))