metadata
base_model: unsloth/phi-4-mini-reasoning
tags:
- text-generation-inference
- transformers
- unsloth
- phi3
license: cc-by-nc-3.0
language:
- en
Uploaded finetuned model
- Developed by: Harsh Srivastava
- License: cc-by-nc-3.0
- Finetuned from model : unsloth/phi-4-mini-reasoning This phi3 model was trained 2x faster with Unsloth and Huggingface's TRL library.
Phi-4 Mini Reasoning – JEE Mathematics Finetuned Model
Developer
Harsh Srivastava
Base Model
unsloth/phi-4-mini-reasoning
Description
This model is a finetuned version of Phi-4 Mini Reasoning designed for solving JEE-level mathematics problems.
The model is optimized for step-by-step mathematical reasoning and symbolic problem solving.
Training Dataset
Total samples used: 356,532 not that much but above 200k samples trained we are still training it better on various datasets for jee by the help of the keyword filters
Sources:
- AI-MO/NuminaMath-TIR — 68,850
- MetaMathQA — 230,808
- TIGER-Lab MathInstruct — 125,220
- PhysicsWallahAI JEE Main 2025 (Jan) — 182
- PhysicsWallahAI JEE Main 2025 (Apr) — 169
- MMLU High School Mathematics — 78
- MMLU College Mathematics — 50
- MMLU Abstract Algebra — 25
Training Details
Base model: Phi-4 Mini Reasoning
Framework: Unsloth + HuggingFace TRL
Training method: LoRA finetuning
Sequence length: 2048
Optimizer: AdamW 8bit
Purpose
The model is designed for:
- JEE mathematics reasoning
- Step-by-step mathematical explanations
- Competitive exam problem solving