Evidence guide · 11 published records

AI for scheduling and resource allocation

Compare bounded tests of shift coverage, queues, routing, staffing, priorities, and constrained assignments.

The decision this guide supports

Can AI satisfy several interacting constraints without hiding an invalid assignment inside a polished table?

Published records
11
Average first score
4/10
Average final score
6.7/10
Average gain
+2.7
Worked / mixed / failed
6 / 0 / 5

Averages use 11 records with a disclosed first score. The set contains 10 synthetic benchmarks and 1 file-backed test; it is a task collection, not a representative model leaderboard.

What the records compare

Evidence before a recommendation.

Coverage, eligibility, ordering, capacity, deadlines, and the smallest safe correction after a failed check.

Across this released set, 5 records failed and 0 remained mixed after the single correction. Those outcomes stay in the guide because a useful decision needs the misses as well as the wins.

  1. Define every hard constraint before prompting.
  2. Validate every row or assignment, not just the totals.
  3. Keep a human owner for exceptions and policy decisions.

Representative evidence

Open the prompts and checks.

The sample deliberately includes different verdicts when available. Every card opens to the full first result, correction, final result, and evidence boundary.