Skip to content
MirrorFirm

What licence questions matter when generated tasks feed a public benchmark?

Buyers should clarify permitted evaluation, training and redistribution uses before placing generated tasks, excerpts or derived artefacts into a public benchmark.

**What uses should we describe when discussing a licence?**

Describe whether the generated material will support internal evaluation, model training, external testing or a published benchmark. Explain whether contractors or other organisations need access. These are questions to settle explicitly rather than permissions to infer from receiving a sample.

**What might publication include beyond the original records?**

A benchmark release can expose task prompts, evidence excerpts, reference answers, agent traces and derived annotations. Ask which of those artefacts may be redistributed, whether adapted versions are covered and how any required attribution should appear.

**Does public access through a website or MCP server settle reuse rights?**

Ask for the applicable licence and usage conditions for the specific material. Technical accessibility and permission to incorporate content into another product are separate questions.

**What else should be clarified before starting?**

Ask how samples, pilot material and later releases are identified, and whether their permitted uses differ. The commercial output is generated material; its grounding and disclosure risk still warrant inspection. These questions define the requested agreement without assuming any particular contract term.

Match the requested licence scope to how your team intends to run, publish and maintain the evaluation.

Audience:
Evaluation teams
Content type:
FAQ
Buyer question:
Scope and getting started

Content release: r-a220eb6e099b243b

More on Ledger adjustment