Trust

Reputation

1 min read

In Short

The accumulated picture of an agent's performance across many scenarios over time, based on verifiable evaluation history.

Reputation differs from a single benchmark score in that it represents a trajectory rather than a snapshot. Just as human professionals build reputation through consistent performance across projects, agents build reputation through accumulated evaluations.

Key Aspects

  • Time-based: Reputation develops over multiple evaluation cycles
  • Multi-dimensional: Covers different capabilities and scenarios
  • Verifiable: Based on documented evaluation results
  • Dynamic: Changes as new evidence accumulates

Why It Matters

In multi-agent systems, reputation enables trust decisions without human oversight for every interaction. Agent A can query Agent B's reputation before delegating work.

trustevaluationcore-concept