metadata
annotations_creators:
- machine-generated
language:
- en
license: mit
multilinguality:
- monolingual
size_categories:
- 1K<n<10K
source_datasets:
- original
tags:
- code-generation
- python
- algorithms
- synthetic-benchmark
- computer-science
task_categories:
- text-generation
- text2text-generation
pretty_name: Algomyth Code Generation Benchmark
dataset_info:
features:
- name: Algorithm Name
dtype: string
- name: Algorithm Type
dtype: string
- name: Time Complexity
dtype: string
- name: Space Complexity
dtype: string
- name: Data Structure
dtype: string
- name: Algorithm Paradigm
dtype: string
- name: Question
dtype: string
- name: Input
dtype: string
- name: Expected Output
dtype: string
- name: Python Code Snippet
dtype: string
splits:
- name: train
num_bytes: 1264865
num_examples: 2000
Algomyth Code Generation Benchmark
Parent GitHub Repository: https://github.com/abdullah00ashraf/abdullah00ashraf
Ecosystem Documentation: Axiyon / Producttolaunch Catalog
Part of the Axiyon Intelligence / Producttolaunch Research Ecosystem.
Author: Abdullah Ashraf (@abdullahashraf122)
GitHub Ecosystem: https://github.com/abdullah00ashraf
Dataset Description
Algomyth is an algorithmic evaluation and synthetic prompt benchmark designed for evaluating LLM code-generation accuracy across searching, sorting, and discrete mathematical algorithms. It features 1,000 standard algorithmic challenge formulations paired with 1,000 varied prompt phrasings, expected complexity boundaries, input test cases, and reference Python implementations.
Features
- Algorithm Name: e.g. Linear Search, Binary Search
- Algorithm Type: Category (Searching, Sorting, Graph, DP)
- Time & Space Complexity: Theoretical Big-O bounds ($O(n)$, $O(1)$)
- Question / Prompt: Natural language user query prompt variations
- Expected Output: Ground-truth evaluation result
- Python Code Snippet: Canonical reference implementation
Usage
from datasets import load_dataset
dataset = load_dataset("abdullahashraf122/algomyth-code-generation-benchmark")
print(dataset["train"][0])