ALPHA-MATH

Math Reasoning Agent/2026•PYTHON / PYTORCH / TRANSFORMERS / DEEPSEEK-MATH-7B / SYMPY / ACCELERATE / KAGGLE

PYTHONPYTORCHTRANSFORMERSDEEPSEEK-MATH-7BSYMPYACCELERATEKAGGLE

Case Study

PROJECT OVERVIEW

/**

* @projectALPHA-MATH

* @descriptionAn offline math-reasoning agent built for AIMO Progress Prize constraints: Kaggle GPU, no external APIs, no internet at inference. DeepSeek-Math-7B-Instruct plans solution steps, a restricted Python/SymPy sandbox executes numeric work, and the agent retries + majority-votes toward an integer answer. System-2 tool loop — not a bare LLM chat wrapper.

* @statusCase Study

Github Kaggle

DISCIPLINECreative Code

ROLEML Engineer

YEAR2026

CATEGORYMath Reasoning Agent

TOOLS

PYTHON
PYTORCH
TRANSFORMERS
DEEPSEEK-MATH-7B
SYMPY
ACCELERATE
KAGGLE

LOGIC BREAKDOWN

PROJECT GALLERY

Preview

alpha-math-reasoning-agent.tsx

>_~/projects/alpha-math-reasoning-agent

main

UTF-8

Prettier: ReadyTypeScript