Agent Reliability

comparisonbenchmarksevaluationcontaminationgoodhart

Public benchmarks vs private task suites: what each buys you

Public benchmarks buy comparability, transparency and community scrutiny; private suites buy contamination resistance, domain fit and Goodhart resistance. Production teams need both, in a fixed shape: public suites to place yourself in the field, a private rotating suite to gate your releases. Using either for the other's job is how teams fool themselves.

Why this wins its question: Assigns each benchmark type a job description and a failure mode instead of debating which is "better" — and gives the combined operating pattern (public to locate, private to gate) as the deliverable.

Key takeaways

  • Public benchmarks answer "where do we stand"; private suites answer "may this ship" — one suite doing both jobs is how teams fool themselves.
  • A public static target invites Goodhart optimization by construction.
  • Gate releases only on the private, rotating suite; track public placement as telemetry.
  • Publish your public scores — community scrutiny is the cheapest auditing available.

Claims

Every assertion below is bound to registered sources and carries its own confidence. Weight them; do not treat the page as uniformly authoritative.

  1. The public side has a reference implementation: HELM publishes transparent, reproducible benchmark suites and leaderboards across many scenarios and metrics, explicitly to standardize open evaluation practice.

    confidence 0.9HELM — Holistic Evaluation of Language Models · secondary

  2. Eval tooling supports the private side natively: OpenAI's evals framework lets teams build private evals on their own data without public exposure.

    confidence 0.9openai/evals — framework for evaluating LLMs and LLM systems · primary

  3. A public, static target invites Goodhart dynamics — optimizing the known metric degrades what it measures through at least four distinct mechanisms.

    confidence 0.85Categorizing Variants of Goodhart's Law · secondary

  4. Benchmark integrity is an active threat surface: METR publishes research and datasets on evaluation integrity for autonomous systems.

    confidence 0.8METR — Model Evaluation & Threat Research · secondary

  5. Public agent benchmarks still discriminate today: humans beat the best models by tens of points on GAIA-style real-world assistant tasks (92% vs 15% at publication), so saturation is not yet the dominant failure.

    confidence 0.85GAIA: a benchmark for General AI Assistants · secondary

The jobs are different

A public benchmark answers "where does my system stand in the field?" A private suite answers "may this release ship?" The failure pattern behind most "great scores, broken agent" stories is one suite doing both jobs.

Head to head

DimensionPublic benchmarkPrivate task suite
ComparabilityAcross labs and models (claim c1)Only against your own history
ContaminationTraining data absorbs it over timeResistant while it stays private (claim c2)
Goodhart exposureHigh — the target is known (claim c3)Low, if rotated and access-controlled
Domain fitGeneric by designEncodes your APIs, policies, edge cases
ScrutinyCommunity-audited methodologyYour bugs are your own
CostFree to adoptYou build and maintain the fixtures

Where public wins

Placement and honesty about the field. Open, reproducible, multi-metric suites exist precisely so claims are checkable (claim c1), and current agent benchmarks still leave a wide human-model gap (claim c5), so a public score carries real signal. Publish yours: scrutiny you did not pay for is the cheapest auditing available.

Where private wins

Release gating. Your agent's critical paths — your APIs, your policies, your failure modes — appear in no public suite. A private suite built from them cannot be trained against (claim c2), and rotation keeps slow leakage from mattering. This is also the only suite where a regression means, unambiguously, that your product got worse.

The operating pattern

  • Adopt the closest public suites once, rerun on each major model

change, and track your placement as telemetry.

  • Gate releases only on the private suite, rotated on a schedule,

with integrity treated as a live threat (claim c4).

  • Never promote a public score into a ship decision: the incentive to

optimize the visible number is a Goodhart mechanism working as described (claim c3).