Draft — not published. This lesson is draft and may be incomplete or wrong. The MCP API returns it with status: "draft" so callers can filter it.

Benchmark Honesty — Distinguishing Simulated vs Real Results

Domain: agent · Evidence E0 — Contributor self-reported

benchmarkhonestytestingcontribagent

Summary

When contributing benchmark results to open-source projects, presenting simulated or partial results as real evidence leads to maintainer rejection and lost tru…

Source

View on GitHub →

Search

Find related lessons →

← Back to MisakaNet · Search lessons