algotune/algotune
Coding
AlgoTune: NeurIPS 2025 benchmark of math/physics/CS problems where the model writes code that matches reference output but runs faster than existing implementations.
Run this task
CLI:
inspect eval inspect_harbor/algotune --model openai/gpt-5Python:
from inspect_ai import eval
from inspect_harbor import algotune
eval(algotune(), model="openai/gpt-5")Dataset information
| Harbor registry | algotune/algotune |
| Inspect task | algotune |
| Latest digest | sha256:69d264f15f717a20841c30fe3e91306925a7df1fd08830b3129d92e1b2c4958e |
| Samples | 154 |
| Paper | arxiv |
| Source | https://github.com/oripress/AlgoTune |
See Task Parameters for the parameter set shared across all Harbor tasks.