metadata
dataset_info:
features:
- name: task_id
dtype: string
- name: prompt
dtype: string
- name: canonical_solution
dtype: string
- name: test
dtype: string
- name: entry_point
dtype: string
- name: difficulty_scale
dtype: string
splits:
- name: test
num_bytes: 156939
num_examples: 151
download_size: 62547
dataset_size: 156939
configs:
- config_name: default
data_files:
- split: test
path: data/test-*
qiskit_humaneval_hard
qiskit_humaneval_hard is a dataset for evaluating LLM's at writing Qiskit code. It contains the same problems as in Qiskit/qiskit_humaneval but is not formatted as a completion exercise. Compared
with Qiskit/qiskit_humaneval the benchmark is more challenging as the LLM needs to decide which python modules to use for solving the problems.
Terms of use
- Terms of use: https://quantum.ibm.com/terms
- Privacy policy: https://quantum.ibm.com/terms/privacy