Treasure Hunt
The sixteen-box game
The playable companion to the TreasureBench suite. Open the formal TreasureBench page.
Treasure Hunt is not a separate benchmark. It is a presentation-only guide to existing public exact fixtures, claims, and limitations. The theme can be removed without changing a result.
Interactive companion No new evidence No external service
A monochrome sixteen-box search map
Sixteen outlined boxes with several dashed search routes ending at different boxes and one X mark.
N
01 02 03 04 05 JavaScript is off. Every module remains complete: open “Exact scope and evidence” to compare all declared arrangements and follow the same claim, study, and data links.
Read the metaphor beside the model
Map fragments mean dispersed private signals. Pooled fragments mean information aggregation. One X means one-answer action compression. Separate digging sites mean diversified action portfolios. Copied maps mean common-source correlation. A minimum crew means threshold-team technology.
The formal terms govern. The companion does not rename group discovery, distinct action coverage, duplication, planner regret, private-baseline regret, recovery-budget attainment, source diversity, communication-induced action compression, recovery budget, action portfolio, or threshold team.
01
Better maps, one shovel hole
Pooled map fragments can improve the shared ranking, while one X still sends every shovel to one place.
Formal term: Information aggregation versus one-answer action compression.
Try a declared arrangement
One shared answer Private clue-following Coordinated action portfolio What the registered model says:
0.383468709731
Exact scope and evidence
DD-000 · DD-C-0004 · DD-C-0005 · DD-C-0006 ·
immutable evidence
Limitation: Exact only for the declared sixteen-box model, signal process, and protocol definitions.
Scroll horizontally to see every column.
No-JavaScript outcome table for Better maps, one shovel hole
Arrangement Registered implication One shared answer 0.383468709731 Private clue-following 0.83222784 Coordinated action portfolio 0.859421246199
02
Split the crew
Repeated digging at the same site duplicates effort; separate sites expand the covered action portfolio.
Formal term: Duplication, distinct action coverage, and group discovery.
Try a declared arrangement
Repeat the top target Follow private signals Allocate the planner portfolio What the registered model says:
0.383468709731
Exact scope and evidence
DD-002 · DD-C-0004 · DD-C-0006 ·
immutable evidence
Limitation: Coverage is valuable only through the declared target-success objective; no universal diversity rule is implied.
Scroll horizontally to see every column.
No-JavaScript outcome table for Split the crew
Arrangement Registered implication Repeat the top target 0.383468709731 Follow private signals 0.83222784 Allocate the planner portfolio 0.859421246199
03
Copied maps
Several maps can look plentiful while all copying one source.
Formal term: Common-source correlation and source diversity.
Try a declared arrangement
All use the common source Mix common and independent sources Use independent sources What the registered model says:
correlated evidence; source count stays one
Exact scope and evidence
DD-008B · DD-C-0057 · DD-C-0058 ·
immutable evidence
Limitation: The exact threshold theorem uses the frozen three-target, homogeneous-source, equal-split model.
Scroll horizontally to see every column.
No-JavaScript outcome table for Copied maps
Arrangement Registered implication All use the common source correlated evidence; source count stays one Mix common and independent sources source diversity can widen the action portfolio Use independent sources higher acquisition cost in the frozen source-choice model
04
One map reader
Give the pooled map to one reader while the rest keep independent routes.
Formal term: Selective audience and the one-reader discovery theorem.
Try a declared arrangement
No shared-map readers One shared-map reader Broadcast to every reader What the registered model says:
optimal when shared accuracy q is below private accuracy p
Exact scope and evidence
DD-012 · DD-C-0059 · DD-C-0060 ·
immutable evidence
Limitation: The theorem is not about unrestricted policies, correlated signals, dynamics, or human attention.
Scroll horizontally to see every column.
No-JavaScript outcome table for One map reader
Arrangement Registered implication No shared-map readers optimal when shared accuracy q is below private accuracy p One shared-map reader uniquely optimal when q is above p Broadcast to every reader can compress action diversity in the frozen role class
05
Minimum digging crew
A site opens only when enough crew members dig there; excess crew crowds out other viable sites.
Formal term: Threshold-team technology and diversified viable action portfolios.
Try a declared arrangement
Threshold team τ = 1 Threshold team τ = 2 Threshold team τ = 4 What the registered model says:
at most 8 viable sites with 8 agents
Exact scope and evidence
DD-016 · DD-C-0071 · DD-C-0072 ·
immutable evidence
Limitation: The capacity statement is exact for deterministic threshold opening with no replication value.
Scroll horizontally to see every column.
No-JavaScript outcome table for Minimum digging crew
Arrangement Registered implication Threshold team τ = 1 at most 8 viable sites with 8 agents Threshold team τ = 2 at most 4 viable sites with 8 agents Threshold team τ = 4 at most 2 viable sites with 8 agents
Return to the formal suite
Open TreasureBench
See DD-010
Download companion metadata
Canonical fixture values come from reproducible run 20260720T202314Z_DD-000_88613408_217c602fa0. Interactive choices select already registered rows; they do not run a model, call a provider, or create a result.