Update README.md

README.md

@@ -1,24 +1,117 @@
 ---
-license: mit
-language:
-- bn
-base_model:
-- Qwen/Qwen3-1.7B
+library_name: transformers
+license: apache-2.0
+base_model: Qwen/Qwen3-1.7B
 pipeline_tag: text-generation
+language:
+  - bn
+  - en
 tags:
-- MathLLM
-- math
-- reasoning
+  - math
+  - bengali
+  - reasoning
+  - sft
+datasets:
+  - dipta007/Ganit
 ---
 
-| Metric                  |   Value |
-|-------------------------|---------|
-| ganit_easy_score        |   30.94 |
-| ganit_medium_score      |   28.22 |
-| ganit_hard_score        |   19.89 |
-| ganit_olympiad_score    |   15.76 |
-| mgsm_score              |   48.80 |
-| msvamp_score            |   64.60 |
-| mean_words              |  253.83 |
-| mean_time_taken         |  152.66 |
-| mean_bengali_percentage |   87.79 |
+# GanitLLM-1.7B_SFT
+
+[![Paper](https://img.shields.io/badge/arXiv-Paper-red)](https://arxiv.org/)
+[![Dataset](https://img.shields.io/badge/HuggingFace-Dataset-yellow)](https://huggingface.co/datasets/dipta007/Ganit)
+[![Models](https://img.shields.io/badge/HuggingFace-Models-orange)](https://huggingface.co/collections/dipta007/ganitllm)
+
+## Highlights
+
+**GanitLLM-1.7B_SFT** is a Bengali mathematical reasoning model trained with Supervised Fine-Tuning on the GANIT dataset. This model serves as the foundation for further RL training (GRPO/CGRPO). Key improvements over the base Qwen3-1.7B model:
+
+- **+33.60 accuracy** on Bn-MGSM benchmark (15.20 → 48.80)
+- **+50.50 accuracy** on Bn-MSVAMP benchmark (14.10 → 64.60)
+- **87.79% Bengali reasoning** (vs 19.64% for base model)
+- **77.5% fewer words** in generated solutions (1124 → 253 words)
+
+> **Note**: This is the SFT-only checkpoint. For best results, use the RL-enhanced versions: [GanitLLM-1.7B_SFT_CGRPO](https://huggingface.co/dipta007/GanitLLM-1.7B_SFT_CGRPO) or [GanitLLM-1.7B_SFT_GRPO](https://huggingface.co/dipta007/GanitLLM-1.7B_SFT_GRPO).
+
+## Model Overview
+
+| Property | Value |
+|----------|-------|
+| **Model Type** | Causal Language Model |
+| **Base Model** | Qwen/Qwen3-1.7B |
+| **Parameters** | 1.7B |
+| **Training** | Supervised Fine-Tuning |
+| **Context Length** | 4,096 tokens |
+| **Language** | Bengali, English |
+
+## Training Details
+
+This model was trained with a single-stage pipeline:
+
+1. **Supervised Fine-Tuning (SFT)**: Trained on GANIT-SFT (~11k examples) to ground reasoning in Bengali
+
+### Training Data
+- **Dataset**: GANIT-SFT (11,023 examples)
+- **Format**: Bengali math problems with chain-of-thought reasoning
+- **Structure**: `<think>` tags for reasoning, `<answer>` tags for final answer
+
+## Quickstart
+
+```python
+from transformers import AutoModelForCausalLM, AutoTokenizer
+
+model_name = "dipta007/GanitLLM-1.7B_SFT"
+
+tokenizer = AutoTokenizer.from_pretrained(model_name)
+model = AutoModelForCausalLM.from_pretrained(
+    model_name,
+    torch_dtype="auto",
+    device_map="auto"
+)
+
+problem = "একটি দোকানে ১২টি আপেল আছে। যদি ৫টি আপেল বিক্রি হয়, তাহলে কতটি আপেল বাকি থাকবে?"
+
+prompt = f"""A conversation takes place between the user and the assistant. The user asks a question, and the assistant solves the problem. Please reason step by step in Bengali, and put your final answer in the <answer> </answer> tags.
+
+Question: {problem}"""
+
+messages = [{"role": "user", "content": prompt}]
+text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
+model_inputs = tokenizer([text], return_tensors="pt").to(model.device)
+
+generated_ids = model.generate(**model_inputs, max_new_tokens=2048, temperature=0.7)
+output_ids = generated_ids[0][len(model_inputs.input_ids[0]):].tolist()
+response = tokenizer.decode(output_ids, skip_special_tokens=True)
+print(response)
+```
+
+### Using vLLM
+
+```bash
+vllm serve dipta007/GanitLLM-1.7B_SFT --max-model-len 4096
+```
+
+## Performance
+
+| Model | Bn-MGSM | Bn-MSVAMP | Avg. Words | Bengali % |
+|-------|---------|-----------|------------|-----------|
+| Qwen3-1.7B (base) | 15.20 | 14.10 | 1124 | 19.64% |
+| **GanitLLM-1.7B_SFT** | **48.80** | **64.60** | **253** | **87.79%** |
+
+## Related Models
+
+| Model | Parameters | Training | Link |
+|-------|------------|----------|------|
+| GanitLLM-1.7B_SFT_CGRPO | 1.7B | SFT + CGRPO | [Link](https://huggingface.co/dipta007/GanitLLM-1.7B_SFT_CGRPO) |
+| GanitLLM-1.7B_SFT_GRPO | 1.7B | SFT + GRPO | [Link](https://huggingface.co/dipta007/GanitLLM-1.7B_SFT_GRPO) |
+| **GanitLLM-1.7B_SFT** | 1.7B | SFT | [Link](https://huggingface.co/dipta007/GanitLLM-1.7B_SFT) |
+| GanitLLM-1.7B_CGRPO | 1.7B | CGRPO | [Link](https://huggingface.co/dipta007/GanitLLM-1.7B_CGRPO) |
+
+## Citation
+
+```bibtex
+will be updated
+```
+
+## License
+
+This model is released under the Apache 2.0 License.