task_53Health care and social assistance

17 verifiers · 13 core · 4 secondary · 17 LLM-graded

TaxonomyHealth care and social assistanceMedical and health services managers

Task prompt

You are helping the urgent-access lead at Meadowbrook Behavioral Health Access Services finish the Friday cutover queue before the early-afternoon scheduling huddle. The EHR migration left a mixed behavioral-health referral batch in the access desk: primary-care referrals, school and obstetric handoffs, ED and observation follow-ups, inpatient discharge follow-ups, and self-referrals. Using the attached access desk reference, referral export, signed care-coordination notes, and slot inventory, determine which clients need escalation before ordinary scheduling, which priority clients can be placed into eligible appointments, which priority clients remain unresolved because no eligible clinical slot can be reserved, and which contact restrictions must be carried forward for outreach. Work only from the attached materials. Treat the access desk reference/capacity rules as controlling for routing, slot eligibility, bridge outreach, payer carve-outs, language/modality requirements, and contact restrictions. Treat a signed care-coordination handoff as authoritative over the raw referral export only when it identifies the same client and is finalized within the 2026-05-29 09:30-11:40 CT batch-review window; ignore draft or unsigned notes. Do not treat raw triage bands as dispositions, do not infer diagnoses, do not recommend medications, and do not invent appointment capacity. Do not use blocked, tentative, held, or QA-release-pending slots. Do not override payer, telehealth, language, guardian/caregiver authority, or collateral-contact restrictions. When facts point to more than one pathway, use the highest-action pathway supported by the attached rules and state any remaining scheduling or contact constraint. Produce exactly two deliverables in `/workspace/output/`: * `access_reconciliation_memo.docx` - a structured memo for the urgent-access lead summarizing the escalation queue, slot reservations, unresolved priority cases, bridge-outreach needs, and protected-contact caveats. * `routing_reconciliation_workbook.xlsx` - a supporting workbook for operations showing client-level routing decisions, slot eligibility/reservation decisions, bridge-outreach flags, and unresolved constraints. Do not create any additional memo, reasoning document, duplicate report, alternate workbook, machine-readable grading artifact, or other extra output file in `/workspace/output/`.

System prompt

The system prompt is not available for this bench. Each harness configures its own agent, and the drop records which model ran and how it was tuned — not the prompt it was given.

What it tests

Whether a signed record is allowed to outrank a raw export, and only under the conditions the reference states: the same client, and finalised inside the batch-review window. A draft or unsigned note is not evidence.

What counts as capacity. Blocked, tentative, held and QA-pending slots are not available, however much of the queue is still waiting.

How models fail

The queue gets cleared and the reasoning does not arrive: placements are made without stating the eligibility rule that permitted them, so a reviewer cannot tell a correct placement from a lucky one.

Underneath that, the eligibility call itself is the verifier fewest trials clear — specialty, payer, language, modality and consent limits are applied loosely, and the edge cases in the routine and waitlist paths are where it shows.

Why it matters

This is patient access. A contact restriction dropped on the way to outreach is a real harm to a real person, and a slot reserved that was never available is a promise the clinic cannot keep.

A task is not fully specified by its written instructions. The rest of the spec is in the seed data the run has to read, and in requirements the prompt never spells out — a run only succeeds if it reconstructs enough of the surface.

Source files

xlsx · docx · pdf

Deliverables

docx · xlsx

Golden answer

withheld from the workspace

Verifiers

13 core · 17 LLM-graded

Which of those 4 sources are authoritative and which are planted distractors is not analysed yet — it needs a per-artifact analysis — authoritative vs distractor, extracted requirements, excerpts, the checks that grade each (`tasks/<id>.json spec_surface`, contract 2.8.0).

See the full package