Benchmark neuromorphic algorithms and systems on shared ground.
NeuroBench is an open-source, community-driven benchmark framework that pairs algorithm correctness with system-aware evaluation for fair and representative comparisons.
Why NeuroBench
Collaborative by design
NeuroBench grows through community submissions across tasks, models, metrics, and frameworks.
Reproducible evaluation
The harness provides a consistent pipeline from data loading to metrics reporting across benchmarks.
Representative reporting
Benchmark outputs pair correctness with efficiency-oriented metrics to improve practical comparability.
Two-track benchmark framework
Algorithm Track
Hardware-independent correctness and complexity
Evaluates benchmark tasks with consistent workload and static metrics, enabling cross-model comparison independent of deployment target.
System Track
Deployment-aware timing and efficiency
Captures real-time and systems-level performance. Benchmarks are defined and baseline submissions are being expanded.
Research Challenges
Build with the community
Contribute benchmarks, submit models, extend metrics, and help shape NeuroBench as an open standard for neuromorphic evaluation.