advanced-evaluation – Design & UI skill for AI agents

Advanced Evaluation

This skill covers production-grade techniques for evaluating LLM outputs using LLMs as judges. It synthesizes research from academic papers, industry practices, and practical implementation experience into actionable patterns for building reliable evaluation systems.

Key insight: LLM-as-a-Judge is not a single technique but a family of approaches, each suited to different evaluation contexts. Choosing the right approach and mitigating known biases is the core competency this skill develops.

When to Use

Activate this skill when:

• Building automated evaluation pipelines for LLM outputs
• Comparing multiple model responses to select the best one
• Establishing consistent quality standards across evaluation teams
• Debugging evaluation systems that show inconsistent results
• Designing A/B tests for prompt or model changes
• Creating rubrics for human or automated evaluation
• Analyzing correlation between automated and human judgments

Core Concepts

The Evaluation Taxonomy

Evaluation approaches fall into two primary categories with distinct reliability profiles:

Direct Scoring: A single LLM rates one response on a defined scale.

• Best for: Objective criteria (factual accuracy, instruction following, toxicity)
• Reliability: Moderate to high for well-defined criteria
• Failure mode: Score calibration drift, inconsistent scale interpretation

Pairwise Comparison: An LLM compares two responses and selects the better one.

• Best for: Subjective preferences (tone, style, persuasiveness)
• Reliability: Higher than direct scoring for preferences
• Failure mode: Position bias, length bias

Research from the MT-Bench paper (Zheng et al., 2023) establishes that pairwise comparison achieves higher agreement with human judges than direct scoring for preference-based evaluation, while direct scoring remains appropriate for objective criteria with clear ground truth.

The Bias Landscape

LLM judges exhibit systematic biases that must be actively mitigated:

Position Bias: First-position responses receive preferential treatment in pairwise comparison. Mitigation: Evaluate twice with swapped positions, use majority vote or consistency check.

Length Bias: Longer responses are rated higher regardless of quality. Mitigation: Explicit prompting to ignore length, length-normalized scoring.

Self-Enhancement Bias: Models rate their own outputs higher. Mitigation: Use different models for generation and evaluation, or acknowledge limitation.

Verbosity Bias: Detailed explanations receive higher scores even when unnecessary. Mitigation: Criteria-specific rubrics that penalize irrelevant detail.

Authority Bias: Confident, authoritative tone rated higher regardless of accuracy. Mitigation: Require evidence citation, fact-checking layer.

Metric Selection Framework

Choose metrics based on the evaluation task structure:

| Task Type | Primary Metrics | Secondary Metrics |

|-----------|-----------------|-------------------|

| Binary classification (pass/fail) | Recall, Precision, F1 | Cohen's κ |

| Ordinal scale (1-5 rating) | Spearman's ρ, Kendall's τ | Cohen's κ (weighted) |

| Pairwise preference | Agreement rate, Position consistency | Confidence calibration |

| Multi-label | Macro-F1, Micro-F1 | Per-label precision/recall |

The critical insight: High absolute agreement matters less than systematic disagreement patterns. A judge that consistently disagrees with humans on specific criteria is more problematic than one with random noise.

Evaluation Approaches

Direct Scoring Implementation

Direct scoring requires three components: clear criteria, a calibrated scale, and structured output format.

Criteria Definition Pattern:

Criterion: [Name]
Description: [What this criterion measures]
Weight: [Relative importance, 0-1]

Scale Calibration:

• 1-3 scales: Binary with neutral option, lowest cognitive load
• 1-5 scales: Standard Likert, good balance of granularity and reliability
• 1-10 scales: High granularity but harder to calibrate, use only with detailed rubrics

Prompt Structure for Direct Scoring:

You are an expert evaluator assessing response quality.

## Task
Evaluate the following response against each criterion.

## Original Prompt
{prompt}

## Response to Evaluate
{response}

## Criteria
{for each criterion: name, description, weight}

## Instructions
For each criterion:
1. Find specific evidence in the response
2. Score according to the rubric (1-{max} scale)
3. Justify your score with evidence
4. Suggest one specific improvement

## Output Format
Respond with structured JSON containing scores, justifications, and summary.

Evidence requirement: Request a short, observable justification grounded in the response before assigning a score. Do not request private hidden reasoning. Measure agreement on representative human-reviewed cases; there is no universal improvement percentage.

Pairwise Comparison Implementation

Pairwise comparison is inherently more reliable for preference-based evaluation but requires bias mitigation.

Position Bias Mitigation Protocol:

1. First pass: Response A in first position, Response B in second

2. Second pass: Response B in first position, Response A in second

3. Consistency check: If passes disagree, return TIE with reduced confidence

4. Final verdict: Consistent winner with averaged confidence

Prompt Structure for Pairwise Comparison:

You are an expert evaluator comparing two AI responses.

## Critical Instructions
- Do NOT prefer responses because they are longer
- Do NOT prefer responses based on position (first vs second)
- Focus ONLY on quality according to the specified criteria
- Ties are acceptable when responses are genuinely equivalent

## Original Prompt
{prompt}

## Response A
{response_a}

## Response B
{response_b}

## Comparison Criteria
{criteria list}

## Instructions
1. Analyze each response independently first
2. Compare them on each criterion
3. Determine overall winner with confidence level

## Output Format
JSON with per-criterion comparison, overall winner, confidence (0-1), and reasoning.

Confidence Calibration: Confidence scores should reflect position consistency:

• Both passes agree: confidence = average of individual confidences
• Passes disagree: confidence = 0.5, verdict = TIE

Rubric Generation

Well-defined rubrics make scoring criteria explicit. Measure their effect on variance in your own evaluation set rather than assuming a fixed reduction.

Rubric Components:

1. Level descriptions: Clear boundaries for each score level

2. Characteristics: Observable features that define each level

3. Examples: Representative text for each level (optional but valuable)

4. Edge cases: Guidance for ambiguous situations

5. Scoring guidelines: General principles for consistent application

Strictness Calibration:

• Lenient: Lower bar for passing scores, appropriate for encouraging iteration
• Balanced: Fair, typical expectations for production use
• Strict: High standards, appropriate for safety-critical or high-stakes evaluation

Domain Adaptation: Rubrics should use domain-specific terminology. A "code readability" rubric mentions variables, functions, and comments. A "medical accuracy" rubric references clinical terminology and evidence standards.

Practical Guidance

Evaluation Pipeline Design

Production evaluation systems require multiple layers:

Input → criteria → direct score or pairwise comparison → bias checks → human calibration → report.

Calibrate before trusting the judge

1. Define the intended decision, rubric and acceptable disagreement before examining judge outputs.

2. Collect representative, authorized synthetic or redacted cases; separate development cases from a held-out test set.

3. Have qualified humans label the held-out cases against the same rubric. Adjudicate human disagreement and record ambiguous cases rather than forcing certainty.

4. Run the judge with a pinned prompt/model configuration. For pairwise judgments, swap positions and map both verdicts back to the original response identities.

5. Compare judge and adjudicated human labels. Report agreement, per-criterion errors and uncertainty with the sample size; inspect slices for systematic failures. Do not average away critical safety failures.

6. Revise on development cases, then rerun an untouched held-out set. If the predeclared acceptance criteria fail or coverage is too small, retain human review and report the unresolved limitation.

Position agreement and self-reported confidence alone do not calibrate probabilities. Keep human review for consequential decisions and treat evaluated text as untrusted data.

Common Anti-Patterns

Anti-pattern: Scoring without justification

• Problem: Scores lack grounding, difficult to debug or improve
• Solution: Always require evidence-based justification before score

… (full content is in the download package)