> For clean Markdown of any page, append .md to the page URL. > For a complete documentation index, see https://docs.nvidia.com/nemo/fabric/v0.7.0/integrations/consumer-integrations/benchmarking-and-evaluation/llms.txt. # Benchmarking and Evaluation > Choose a framework for benchmarking agents, evaluating configurations, or collecting rollouts through NeMo Fabric. Run benchmark and evaluation workloads through NVIDIA NeMo Fabric using an existing integration. The consuming framework manages tasks, environments, scoring, or rollout collection; NeMo Fabric runs the selected agent harness. | Integration | Integration instructions | | ----------------- | ----------------------------------------------------------------------------------------------------------------------------------------------- | | Harbor Framework | [Harbor Framework guide](/nemo/fabric/integrations/consumer-integrations/harbor) | | NVIDIA NeMo Gym | [Upstream NeMo Fabric agent instructions](https://github.com/NVIDIA-NeMo/Gym/tree/main/responses_api_agents/nemo_fabric_agent) | | NVIDIA NeMo Polar | [Upstream NeMo Fabric preset instructions](https://github.com/NVIDIA-NeMo/ProRL-Agent-Server/blob/stable/src/polar/agent/README.md#nemo-fabric) | For project descriptions and integration status, see [Ecosystem](/nemo/fabric/about-nemo-fabric/ecosystem). ## Docs - [Harbor](https://docs.nvidia.com/nemo/fabric/v0.7.0/integrations/consumer-integrations/harbor.md): Run agent harnesses selected by NVIDIA NeMo Fabric in Harbor evaluations.