A subjective criterion requires the exact heading or label your golden happens to use, words no prompt contains.
Do this instead
Name the element by what it is, and grade how it is presented, not what it is called.
Where it happened
One real task. Each shows the spot and what it should have been; click any line to open the task there.
the words at faultthe row or field at faultmissing from the taskwhat is righta task file
Example 32Small Business & Personal Docs · 3 Leg A turns
What is wrongNine of ten subjective criteria require the golden page's own headings and labels, such as “Claim 1 - May trip”, which none of the three Leg A prompts asks for.
In reimbursement_status.html, the determination table's header row is set apart from its data rows by colour or weight.
How it breaks the standardNo Leg A prompt asks for these headings; they are the golden's choices. Each criterion copies the golden page, so nothing flags it, but a correct page with its own headings fails subjective criteria 1 and 4 and seven more. Only subjective criterion 2 names a property without quoting text.
Name ONE observable property of the RENDERED artifact. The judge sees the render, not the intent behind it.
A property, not one build's wording.
Guidelines · 8.1 Writing Subjective Rubrics
If anything related to the render is asked in the prompt, it requires an objective rubric ONLY.
A label you chose is graded nowhere.
Why it matters
Subjective rubrics grade how the artifact reads: layout, hierarchy, spacing, consistency. A criterion naming one build's text rejects a subset of valid implementations, and a second correct run presenting the same results under its own wording fails almost the whole block. Overfitting counts toward the share of criteria built wrong, and the task fails once that share passes 15% of all its criteria. Where most of the block is written this way, rewriting three or four is not enough.
In `reimbursement_status.html`, under `The determination (as of 28 August 2026)`, a four-column table has the header row Claim, Reached your account?, Due by 28 Aug?, and Verdict.
It's the 28th of August and I'm trying to close out something that's been sitting in the back of my head.
I drove my own car to Everett twice this year for work, once in May and once in August, and I filed a mileage claim for each trip. Marguerite approved both of them and HR wrote back saying they were acknowledged and complete and that the money comes through payroll, no separate deposit, nothing else needed from me. Which is fine, except my paycheck is the same number every single time and I honestly can't tell whether either one of those ever actually landed.
I put some things in inputs for you, the expense policy and a couple of my pay statements, plus something the bank sent me. My mail and my accounts should have the rest.
Can you work out, for each of the two claims separately, whether the money reached me or not? I don't want them added together into one number, I want to know which one is which. And if one of them didn't come through, tell me whether I can still do anything about it, because I have a feeling the policy puts a clock on that.
Put it together as reimbursement_status.html. I want to be able to open it and read it properly before I send anything to Payroll.
Leg A turn 2
I found another pay statement and attached it as IMG_4471.jpg. Please check it against the policy, the claim emails, and the account records, then update the same reimbursement_status.html with whatever it changes. Keep the two claims separate and preserve everything that remains supported.
Leg A turn 3
Before I send this, for any claim you determine is overdue and unpaid, put the discrepancy deadline on the page as an exact date and say whether it had already passed as of the 28th. Keep using reimbursement_status.html, keep the two claims separate, and preserve everything that remains supported.
From the taskSubjective criterion 2
2
In `reimbursement_status.html`, each claim card carries an uppercase tinted status pill immediately under the claim heading.