metadata
license: mit
task_categories:
- question-answering
language:
- en
tags:
- code
size_categories:
- 1K<n<10K
Introduction: >-
This is the AlgoSimBench dataset, which contains competitive-level programming
problems and fine-grained algorithm tags for topics related to the problems
and corresponding algorithms to solve them. The purpose of AlgoSimBench is to
evaluate models' abilities in identifying algorithmically similar problems.
AlgoSimBench-mcq consists of 402 multiple-choice questions, with a reference
problem, one algorithmically similar positive and 3 algorithmically dissimilar
negatives. Each problem_id corresponds to ones in the raw problem corpus.
configs:
- config_name: default
data_files:
- split: MCQ_test
path: data/MCQ_test-*
dataset_info:
features:
- name: negative
list:
- name: prob_id
dtype: string
- name: tags
sequence: string
- name: positive
struct:
- name: prob_id
dtype: string
- name: tags
sequence: string
- name: reference
struct:
- name: prob_id
dtype: string
- name: tags
sequence: string
splits:
- name: MCQ_test
num_bytes: 141382
num_examples: 402
download_size: 31573
dataset_size: 141382