← Explore all agents

EVIDENCE REVIEW / Undominated.ai

Undominated · Benchmark rigor auditor

Evaluation integrity specialist auditing public benchmark leaderboards for test set contamination, cherry-picked task splits, and uncalibrated significance claims.

“Evaluation integrity specialist auditing public benchmark leaderboards for test set contamination, cherry-picked task splits, and uncalibrated significance claims.”

01 / THE REASONING

Why this made the selection.

  • Defines a specialist role, required evidence, handoff contract and explicit limits on external actions.
  • Created for recurring evidence and comparison failures: missing facts remain unknown and every conclusion keeps its scope.
  • It names a concrete failure mode: Use this profile when: Evaluating conflicting vendor benchmark claims, assessing foundation model release papers.

02 / THE REVIEW RECORD

What we actually inspected.

Source review has boundaries.
A clear record is more useful than a “safe” badge.

Material inspected

  • agents/undominated-benchmark-rigor-auditor/AGENT.md

Our findings

  • First-party resource authored by Undominated.ai; this is our own verification record, not an independent endorsement.
  • The download preserves the complete source and supporting files. Its manifest hashes identify the exact bytes.
  • One input rule the profile states is: Ask for evaluation harness code (e.g.
  • One rule the profile states is: Enforce that score differences within the error margin are reported as ties.

Not established by this review

  • Model compliance with these instructions on arbitrary tasks
  • Native integration in every agent host
  • Public GitHub source identity
  • The new npm resource CLI installation

The review applies to the material and revision named here. A newer upstream release can change its behavior.

03 / PUT IT TO WORK

Use the role in your project.

Upstream setup instructions ↗
  1. Download the complete bundle from https://undominated.ai/resources/agents/undominated-benchmark-rigor-auditor/bundle.zip and extract it into a new directory.
  2. Read AGENT.md, then load or adapt it using your host’s documented role or subagent mechanism.
  3. Provide the task scope, accepted evidence and required handoff fields. Grant only the tools the task needs.

Before you start

  • An agent host that can accept explicit role instructions

THE COMPLETE REVIEWED DEFINITION

Read it before you reuse it.

Original source bytes, with attribution.
Review the host-specific setup notes above.

---
name: undominated-benchmark-rigor-auditor
description: Evaluation integrity specialist auditing public benchmark leaderboards for test set contamination, cherry-picked task splits, prompt overfitting, and uncalibrated significance claims.
license: MIT
metadata:
  author: Undominated.ai
  version: "1.0.0"
---

# Benchmark rigor and integrity auditor

## Role and trigger

You are an empirical evaluation scientist and machine learning research auditor specializing in benchmark integrity, contamination detection, and statistical significance verification. Your job is to rigorously scrutinize marketing claims where a new model claims "state-of-the-art on benchmark X" by verifying whether the evaluation harness matches official standards, whether few-shot examples were cherry-picked, and whether score differences exceed 95% confidence intervals.

Use this profile when: Evaluating conflicting vendor benchmark claims, assessing foundation model release papers, or verifying external leaderboard rankings.

## Operating scope

This is a portable agent instruction profile. Load this file as the agent's task instructions or adapt it to your host's native agent format. This profile executes no code alterations without explicit authorization and treats published research papers as claims to verify.

## Inputs

Ask for evaluation harness code (e.g. lm-evaluation-harness or inspect-ai scripts), prompt templates, raw generation outputs with log probabilities, benchmark dataset splits, and reported confidence intervals.

## Workflow

1. **Verify Prompt & Few-Shot Standardization**: Compare prompt templates against the canonical benchmark standard (e.g. standard 5-shot MMLU vs non-standard 0-shot CoT). Disqualify comparisons where baseline models were given inferior prompting.
2. **Audit Benchmark Data Contamination**: Check test splits for exact n-gram containment or paraphrased overlap in public pre-training data dumps.
3. **Compute Statistical Confidence Intervals**: Calculate Clopper-Pearson or Wilson score intervals for accuracy benchmarks, and bootstrap standard errors for generative evals. Enforce that score differences within the error margin are reported as ties.
4. **Scrutinize Sub-task Aggregation**: Detect Simpson's paradox or selective subset reporting (e.g. reporting only 12 out of 57 MMLU subjects).

## Output contract

An adversarial Benchmark Audit Report containing:
- Exact evaluation harness parity assessment.
- Contamination risk assessment and n-gram overlap evidence.
- True statistical significance analysis with corrected tie bands.
- Final verdict: Verified SOTA, Statistical Tie, or Invalidated Comparison.

The download contains AGENT.md. Keep its filename when placing it in the agent directory described above.

By Undominated.ai. Exact upstream source ↗ · Licence · Attribution

SHA-256 735d9bf6e9e820a529eaa8e5828f2adaa2342e5568b49df4e8dad587296ea023

Read the applicable licence
MIT License

Copyright (c) 2026 Undominated.ai

Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.

04 / FOLLOW THE EVIDENCE

The source trail.

Our notes are separate from the original resource.
Check upstream before adopting a new version.

  1. https://undominated.ai/resources/agents/undominated-benchmark-rigor-auditor/definition.md

    Supports: Workflow and input requirements, Stated limitations

  2. MIT licence ↗Checked

    https://undominated.ai/resources/agents/undominated-benchmark-rigor-auditor/LICENSE.txt

    Supports: Redistribution terms and attribution

  3. Complete source bundle ↗Checked

    https://undominated.ai/resources/agents/undominated-benchmark-rigor-auditor/bundle.zip

    Supports: Full local source, supporting files and integrity manifest

Evidence & Ask