PUBLICATION REVISION 9

Group assignment benchmark cases

Results for 14 fixed multi-session assignment cases. Each report lists the input size, tool settings, individual runs, validator scores, and downloadable files.

Case reports

Select a case for per-tool and per-attempt details.

14 reports

Published benchmark case reports
CaseTypeDimensionsLimitDescription
Kirkman Schoolgirls 15×5×7public.kirkman-schoolgirls-15x5x7.v1Repeated mixing15 people
5 groups × 7 sessions
60s
8 CPUs
The classical problem of arranging fifteen participants into five groups over seven sessions while minimizing repeated pairs.
Three golfer foursome rotationsinternal.reddit-golfers-28p-7g-3s.v1Repeated mixing28 people
7 groups × 3 sessions
60s
8 CPUs
A three-round synthetic reconstruction for 28 golfers in seven foursomes.
Social Golfer 32×8×10public.social-golfer-32x8x10.v1Repeated mixing32 people
8 groups × 10 sessions
60s
8 CPUs
Thirty-two participants rotate through eight groups for ten sessions with unique contacts as the canonical objective.
Social Golfer 32x8x15 (8-4-15)internal.social-golfer-32x8x15.v1Repeated mixing32 people
8 groups × 15 sessions
60s
8 CPUs
Pure Social Golfer overbound repeat-pressure stretch: 32 participants into 8 groups of 4 across 15 sessions with a meet-at-most-once target.
Social Golfer 9–5–10 (45×9×10)public.social-golfer-45x9x10.v1Repeated mixing45 people
9 groups × 10 sessions
60s
8 CPUs
Forty-five participants rotate through nine groups of five for ten sessions; a checked construction attains the proven objective minimum of zero repeated pairs for this fixed ten-session problem.
Six student workshop rotationsinternal.real-world-workshops-120x12x6.v1Repeated mixing120 people
12 groups × 6 sessions
60s
8 CPUs
A 120-person six-workshop-session no-repeat reconstruction.
Small constrained workshopexample.small-workshop-constrained.v1Constrained assignment8 people
2 groups × 3 sessions
60s
8 CPUs
A representative small workshop combining hard and soft grouping constraints across multiple sessions.
Scout patrol leader rotationinternal.stack-scout-patrol-rota-8p-4g-7s.v1Constrained assignment8 people
4 groups × 7 sessions
60s
8 CPUs
An eight-person patrol rota combining exact pair pressure, patrol visits, and three-session spacing.
Canonical station rotation 18×6×6internal.station-rotation-18p-6g-6s.v1Constrained assignment18 people
6 groups × 6 sessions
60s
8 CPUs
Eighteen participants rotate through six named stations while squared penalties discourage missed station visits and repeated encounters.
Anonymized 57-person sailing tripinternal.sailing-trip-57p-5g-5s.v1Constrained assignment57 people
5 groups × 5 sessions
60s
8 CPUs
An owner-authorized, anonymized five-session sailing workload preserving the complete objective and mixed-constraint structure of the recovered GroupMixer setup.
Partial-attendance keep-apart capacity pressure, 152 participantsinternal.partial-attendance-keep-apart-capacity-pressure-152p.v1Large/capability case152 people
12 groups × 6 sessions
300s
8 CPUs
The deterministic partial-attendance capacity workload extended with dense session-scoped hard separation windows.
Social Golfer 196x14x14 (14-14-14)internal.social-golfer-196x14x14.v1Large/capability case196 people
14 groups × 14 sessions
300s
8 CPUs
Synthetic pure Social Golfer scenario: 196 participants into 14 groups of 14 across 14 sessions, maximizing unique contacts.
Minimum-occupancy rotation, 280 participants, 36 groups, 41 sessionsinternal.minimum-occupancy-rotation-280p-36g-41s.v1Large/capability case280 people
36 groups × 41 sessions
300s
8 CPUs
A 280-participant rotation through 36 groups of capacity 8 and minimum occupancy 6 across 41 sessions, minimizing repeated encounters.
Hard-infeasible three-person separation trianglepublic.hard-infeasible-apart-triangle-3p-2g-1s.v1Proof behavior3 people
2 groups × 1 sessions
60s
8 CPUs
Three participants must be pairwise separated in one session with only two groups, creating a small well-formed hard-feasibility contradiction.

Runs

Randomized settings are run with five fixed seeds. Deterministic settings are run once.

Validation

Schedules are scored only after independent validation against the canonical case.

Comparability

Scores are comparable within one case, not between different cases. Run times from hosted websites are not compared with programs run on the benchmark machine.