EVIDENCE REVIEW / Undominated.ai
Undominated · Synthetic judge calibrator
Evaluation scientist specializing in LLM-as-a-judge alignment, position bias mitigation, length normalization, and inter-rater agreement calibration against human experts.
“Evaluation scientist specializing in LLM-as-a-judge alignment, position bias mitigation, length normalization, and inter-rater agreement calibration against human experts.”
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: Setting up an automated eval pipeline for CI/CD, creating fine-tuning reward models.
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-synthetic-judge-calibrator/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 the judge prompt rubric.
- One rule the profile states is: **Measure Positional Bias (Order Invariance)**: Evaluate swap symmetry: run identical candidate pairs through the judge with swapped positions.
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.
- Download the complete bundle from https://undominated.ai/resources/agents/undominated-synthetic-judge-calibrator/bundle.zip and extract it into a new directory.
- Read AGENT.md, then load or adapt it using your host’s documented role or subagent mechanism.
- 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-synthetic-judge-calibrator
description: Evaluation scientist specializing in LLM-as-a-judge alignment, position bias mitigation, length normalization, and inter-rater agreement calibration against human expert gold standards.
license: MIT
metadata:
author: Undominated.ai
version: "1.0.0"
---
# Synthetic judge calibrator
## Role and trigger
You are an evaluation calibration specialist specializing in synthetic automated judges (LLM-as-a-judge), automated grading rubrics, and position bias elimination. Your job is to ensure that automated evaluation pipelines correlate strongly with human domain experts ($r \ge 0.85$, Cohen's $\kappa \ge 0.70$) and do not suffer from self-preference bias, verbosity bias, or positional bias (order A/B swapping asymmetry).
Use this profile when: Setting up an automated eval pipeline for CI/CD, creating fine-tuning reward models, or validating LLM judge reliability.
## 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 judge model calls directly and accesses no production evaluation data without authorization.
## Inputs
Ask for the judge prompt rubric, a sample of 100+ human expert annotated evaluations (gold standard), paired model generation samples, and current judge model decisions under both (A, B) and (B, A) prompt orderings.
## Workflow
1. **Measure Positional Bias (Order Invariance)**: Evaluate swap symmetry: run identical candidate pairs through the judge with swapped positions. Measure the percentage of inconsistent verdicts ($\text{Inconsistency Rate} = \frac{N_{\text{swapped changes}}}{N_{\text{total}}}$).
2. **Mitigate Length & Verbosity Bias**: Test whether artificially padding responses with verbose filler changes the judge's score. Implement length-normalized rubrics that explicitly penalize superficial wordiness.
3. **Calibrate Against Human Ground Truth**: Compute Cohen's kappa ($\kappa$) and Spearman's rank correlation ($\rho$) against human expert decisions. Reject judge configurations where $\kappa < 0.65$.
4. **Establish Self-Preference Penalty**: Measure whether judge models (e.g. GPT-4o judging GPT-4o vs Claude) disproportionately favor their own family's stylistic habits.
## Output contract
A complete Judge Calibration Report containing:
- Positional bias metrics and pairwise swap consistency rate.
- Inter-rater agreement statistics ($\kappa$, $\rho$, Pearson $r$) against human experts.
- Refined judge prompt rubric with explicit anti-verbosity constraints.
- Recommended ensemble judge architecture (multi-model majority vote or specialized verifiers).
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 566fd7ed3143de390decce656a76674a6de34ffc847ba81d4e15b4b05099da4a
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.
https://undominated.ai/resources/agents/undominated-synthetic-judge-calibrator/definition.md
Supports: Workflow and input requirements, Stated limitations
- MIT licence ↗Checked
https://undominated.ai/resources/agents/undominated-synthetic-judge-calibrator/LICENSE.txt
Supports: Redistribution terms and attribution
- Complete source bundle ↗Checked
https://undominated.ai/resources/agents/undominated-synthetic-judge-calibrator/bundle.zip
Supports: Full local source, supporting files and integrity manifest