Evaluations and datasets · Lesson 12 of 18 · 9 min

Find annotation disagreement

Annotation disagreement can reveal an ambiguous rubric rather than a model defect. Resolve the criterion and preserve adjudication notes before using the label as ground truth.

Open the lab Course map

Mastery 0/3Run itExplain itApply it

learner@clairos:~ $

↑ history · Tab complete · Ctrl+C cancel · Esc, Tab to leaveExit 0

Your objectives

Hints are always available. Wrong turns never cost XP.

1
Read the case evidence
Read /opt/ai-lab/cases/aiops-evals-datasets-012.json. Separate the recorded facts from your assumptions.
Hint & worked example
Annotation disagreement can reveal an ambiguous rubric rather than a model defect. Resolve the criterion and preserve adjudication notes before using the label as ground truth.
The complete case JSON is read successfully.
```
cat /opt/ai-lab/cases/aiops-evals-datasets-012.json
```
Common mistake to avoid: Always discard the stricter reviewer without checking the criterion.
2
Isolate the deciding signal
Select the rubric_ambiguous field from the case JSON. What does it establish—and what does it not establish?
Hint & worked example
Clarify and adjudicate the ambiguous rubric before treating either label as definitive.
A successful JSON query isolates rubric_ambiguous.
```
jq .rubric_ambiguous /opt/ai-lab/cases/aiops-evals-datasets-012.json
```
Common mistake to avoid: Using a different case or an unrelated signal as the justification.
3
Record your operational decision
Choose the justified code: clarify-rubric, change-without-evidence, or ignore-the-signal. Save that one code followed by a newline in /home/learner/lab/aiops-evals-datasets-012-decision.txt.
Hint & worked example
Clarify and adjudicate the ambiguous rubric before treating either label as definitive.
The decision artifact contains the correct code, not just a matching command in history.
```
printf '%s\n' 'clarify-rubric' > /home/learner/lab/aiops-evals-datasets-012-decision.txt
```
Common mistake to avoid: Always discard the stricter reviewer without checking the criterion.

Understand the pattern

Annotation disagreement can reveal an ambiguous rubric rather than a model defect. Resolve the criterion and preserve adjudication notes before using the label as ground truth.

Interpret the evidence

Clarify and adjudicate the ambiguous rubric before treating either label as definitive.

A tempting mistake

Always discard the stricter reviewer without checking the criterion.

All names, measurements, prices, versions, and thresholds in this exercise are fictional training data. The evidence supports this bounded scenario, not a claim about a live provider.

Worked example and expected result

$ cat /opt/ai-lab/cases/aiops-evals-datasets-012.json

{
  "fixture": "Fictional training data; not a live measurement.",
  "case": "case-009",
  "reviewer_a": "pass",
  "reviewer_b": "fail",
  "rubric_ambiguous": true
}

Exit status: 0. The command completed successfully in the training fixture.

$ jq .rubric_ambiguous /opt/ai-lab/cases/aiops-evals-datasets-012.json

true

Exit status: 0. The command completed successfully in the training fixture.

$ printf '%s\n' 'clarify-rubric' > /home/learner/lab/aiops-evals-datasets-012-decision.txt

(no output; exit 0)

Exit status: 0. The command completed successfully in the training fixture.

Try a variation

Name the assumption that would have to change for “clarify-rubric” to stop being the justified action. Do not silently replace missing evidence with a guess.

Explain it

What does the evidence in “Find annotation disagreement” justify?Always discard the stricter reviewer without checking the criterion.Make the change now and gather the missing evidence afterward.Clarify and adjudicate the ambiguous rubric before treating either label as definitive.

Apply it elsewhere

A colleague concludes: “Always discard the stricter reviewer without checking the criterion.” What should you do?Treat this fixture as a live measurement.Reject that conclusion; return to the case evidence and its stated limits.Accept it because the training command completed.

Keep the momentum.

Complete the lab, then explain and apply the pattern.

Next: Detect evaluation leakage Try an incident drill

Anonymous lesson and mastery progress lasts for this session. Sign in to keep synced lesson XP between sessions; mastery stays device-local. No speed bonus, lost lives, or hint penalties.

Keep learning

Case methodology: evidence, limits, decision