Connect tasks, environments, and human review.
Comet supports task creation, environment setup, human review, and model evaluation. Its data series cover knowledge work, domain expertise, enterprise environments, and visual reasoning.
A task connects the objective, reference material, starting state, expected deliverable, and evaluation criteria. Teams can inspect agent runs and compare outcomes against the same requirements.
Start with work that is worth evaluating
Explore GDPVal-style knowledge work, domain-specific tasks, and visual reasoning. Each task links an objective to source material and review criteria.
- Knowledge-work tasks and supporting references.
- Domain-specific criteria shaped around your work.
- Visual reasoning over charts, figures, and documents.
Give agents repeatable conditions
Save an application’s initial state and reuse it for task runs. Review the execution trace and final result against the task’s success criteria.
- Document the initial application state.
- Connect task objectives with available context.
- Inspect outcomes using explicit review criteria.
What a program can include
Scope can include:
- Task objectives and reference-material requirements
- Environment starting conditions where relevant
- Task-specific review and evaluation criteria
- Execution traces and review feedback
Illustrative task / Not a production record
Make the objective concrete.
- Objective
- Review a project proposal against its requirements.
- Context
- The proposal, requirements, expected deliverable, and a defined browser starting state.
- Review
- Identify gaps, make recommendations, and support the review with source references. This example is illustrative, not a recorded model result.
Common questions
How is Comet related to Rise Data Labs?
Comet is a Rise Data Labs product that connects expert-built tasks, repeatable environments, and human review for AI agents. Explore its data series to find the workflow and task coverage that fit your model.
How can we discuss a Comet project?
Tell the Rise Data Labs team about your domain, tasks, and review criteria. We’ll confirm the scope and availability of relevant data series and samples.