DAGGER-12B-SFT
Model Description
DAGGER-12B-SFT is a supervised fine-tuned model for computational graph generation in Bangla mathematical reasoning. This is the SFT-only variant, serving as both a standalone model and initialization for GRPO training.
Highlights
- SFT-only training on 3,000 verified computational graph examples
- Strong baseline performance for distractor-aware reasoning
- Foundation for GRPO: Used as initialization for dagger-12B_SFT_GRPO
- Efficient inference: ~334 tokens per problem (weighted average)
Model Overview
| Attribute | Value |
|---|---|
| Base Model | Gemma-3-12B-Instruct |
| Training | Supervised Fine-Tuning |
| Parameters | 12B |
| LoRA Rank | 64 |
| Max Sequence Length | 4096 |
Performance
| Dataset | Original | +Distractor | Drop |
|---|---|---|---|
| MGSM | 70.0 | 56.8 | 13.2 |
| MSVAMP | 76.8 | 65.4 | 11.5 |
| Weighted Avg | - | - | 66.7 |
Comparison with GRPO
| Model | Weighted Avg Accuracy |
|---|---|
| dagger-12B_SFT | 66.7 |
| dagger-12B_SFT_GRPO | 69.4 (+2.7) |
GRPO provides +2.7 points improvement over SFT alone.
Quickstart
from transformers import AutoModelForCausalLM, AutoTokenizer
model_name = "dipta007/dagger-12B_SFT"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(
model_name,
torch_dtype="auto",
device_map="auto"
)
USER_PROMPT_TEMPLATE = """You are an expert Bengali Math Reasoner. Your task is to solve mathematical problems by constructing a "Computational Graph".
### Graph Rules:
- `id`: Unique identifier (e.g., "n1", "n2").
- `val`: The raw number extracted from text (for input nodes).
- `op`: The operation (`add`, `sub`, `mul`, `div`, `round`, `sqrt`, `floor`, `sum`, `mean`). Use `const` for input numbers.
- `args`: List of input node IDs.
- `distractor`: Boolean (`true` / `false`). Set to `true` if the node is NOT used in the final calculation path.
- `label`: Label for the node.
### Available Operations:
- Input: `const` (Use this for all numbers found in text or constants).
- Arithmetic: `add`, `sub`, `mul`, `div`, `abs` (absolute difference).
- Logic/Stats: `sum`, `mean`, `min` (minimum), `max` (maximum).
- Rounding: `round` (nearest int), `floor` (round down), `ceil` (round up).
- Advanced: `sqrt`, `pow`, `mod` (remainder), `gcd`, `lcm`.
- Output: `identity` ("final_result" points to the answer node)
Only output a JSON graph representing the solution, nothing else. Nodes must be topologically sorted, and there must be exactly one "final_result" node that represents the final answer. One example is provided below.
### Example:
Question:
মিনার কাছে ১২২১৯৫ টা কলম আছে। রাজুর কাছে ২৫০৮৪ টা কলম আছে। মিনা রাজুর কাছে ১১২৬ টি কলম চাইল। রাজু ১০০০ টি কলম দিতে রাজি হল, কিন্তু পরে আর দিলেনা। প্রতিটি কলমের দাম ৪৫.৬ টাকা। মিনা যদি কলমগুলো বিক্রি করতে চায়, সে কত টাকা পাবে?
Output:
```json
{{
"nodes": [
{{"id": "n1", "op": "const", "val": 122195, "distractor": false, "label": "মিনার কলম"}},
{{"id": "n2", "op": "const", "val": 25084, "distractor": true, "label": "রাজুর কলম"}},
{{"id": "n3", "op": "const", "val": 1126, "distractor": true, "label": "মিনা রাজুর কাছে চাইল"}},
{{"id": "n4", "op": "const", "val": 1000, "distractor": true, "label": "রাজু দিতে রাজি হল"}},
{{"id": "n5", "op": "const", "val": 45.6, "distractor": false, "label": "প্রতিটি কলমের দাম"}},
{{"id": "total_money", "op": "mul", "args": ["n1", "n5"], "distractor": false, "label": "মিনার মোট টাকা"}},
{{"id": "final_result", "op": "identity", "args": ["total_money"], "distractor": false, "label": "চূড়ান্ত উত্তর"}}
]
}}```
### Your Task:
Question:
{question}
Output:
"""
question = "রজারের 5টি টেনিস বল আছে। সে আরও 2 ক্যান টেনিস বল কিনেছে। প্রতিটি ক্যানে 3টি করে টেনিস বল আছে। তার কাছে এখন কতগুলি টেনিস বল আছে?"
prompt = USER_PROMPT_TEMPLATE.format(question=question)
messages = [
{"role": "user", "content": prompt}
]
text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
inputs = tokenizer(text, return_tensors="pt").to(model.device)
# Generate
outputs = model.generate(**inputs, max_new_tokens=1024, temperature=0.7, top_p=0.8)
response = tokenizer.decode(outputs[0][len(inputs.input_ids[0]):], skip_special_tokens=True)
print(response)
Training Configuration
| Parameter | Value |
|---|---|
| LoRA Rank / Alpha | 64 / 128 |
| Global Batch Size | 256 |
| Epochs | 4 |
| Learning Rate | 1e-5 → 1e-6 |
| Optimizer | AdamW |
| Weight Decay | 0.001 |
| Precision | BF16 |
When to Use This Model
- As a baseline: Compare against GRPO-enhanced variants
- For GRPO initialization: Use as starting point for policy optimization
- Resource-constrained settings: When GRPO training is not feasible
- Research: Studying the effect of SFT vs. GRPO on graph generation
Related Models
| Model | Training | Performance |
|---|---|---|
| dagger-12B_SFT | SFT | 66.7 |
| dagger-12B_SFT_GRPO | SFT → GRPO | 69.4 |
| dagger-12B_GRPO | Base → GRPO | 61.5 |
License and Data Provenance
Model weights are released under the Gemma Terms of Use.
Training data is not fully permissive. Part of the SFT data and all GRPO prompts come
from numina-math-cot-bn, which is CC BY-NC-SA 4.0 (NonCommercial, ShareAlike). For
commercial use, re-derive that portion from the Apache-2.0 upstream
AI-MO/NuminaMath-CoT.
Citation
@inproceedings{nazi2026dagger,
title={{\dag}DAGGER: Distractor-Aware Graph Generation for Executable Reasoning in Math Problems},
author={Zabir Al Nazi and Shubhashis Roy Dipta and Sudipta Kar},
booktitle={Findings of the Association for Computational Linguistics: EMNLP 2026},
year={2026},
eprint={2601.06853},
archivePrefix={arXiv},
primaryClass={cs.CL},
url={https://arxiv.org/abs/2601.06853},
}