Demonstrating automatic discovery of prompting strategies for large integer multiplication
Matt Mazur posed a challenge: get LLMs to accurately multiply integers in the 1-10,000 range through prompting strategies alone (not tool calling). His goal is to discover what prompting techniques actually work.
| # train_grpo.py | |
| # | |
| # See https://github.com/willccbb/verifiers for ongoing developments | |
| # | |
| """ | |
| citation: | |
| @misc{brown2025grpodemo, | |
| title={Granular Format Rewards for Eliciting Mathematical Reasoning Capabilities in Small Language Models}, | |
| author={Brown, William}, |
| import re | |
| from txtai import Embeddings, LLM | |
| # Prompt courtesy of the following link: https://github.com/codelion/optillm/blob/main/cot_reflection.py | |
| def cot(system, user): | |
| system = f""" | |
| {system} | |
| You are an AI assistant that uses a Chain of Thought (CoT) approach with reflection to answer queries. Follow these steps: |
| #!/bin/bash | |
| set -euo pipefail | |
| # Directory setup | |
| APP_DIR="${HOME}/Applications" | |
| ICON_DIR="${HOME}/.local/share/icons" | |
| DESKTOP_DIR="${HOME}/.local/share/applications" | |
| BIN_DIR="${HOME}/.local/bin" |
| # Tested regorously on multiple python environment and multiple devices | |
| # Feel free to update | |
| import numpy as np | |
| from torch.utils.data import Dataset, DataLoader | |
| import random | |
| import torch | |
| import os | |
| from functools import partial |
