INDEPENDENT BENCHMARK PROJECT

Project overview

Workbench online

Build ten samples that genuinely require procedural world state.

Track evidence, annotations, probe behavior, and human decisions without mixing model proposals with benchmark truth.

Open model test
Total slots10Initial feasibility set
Accepted0Passed human audit
In progress0Annotation or probes
Rejected0Retained as lessons
DATA WORKFLOW

Initial sample set

10 slots
SampleDomainMechanismTaskStatusUpdated
S0001Question not draftedUnassigned——candidate—AnnotateAudit
S0002Question not draftedUnassigned——candidate—AnnotateAudit
S0003Question not draftedUnassigned——candidate—AnnotateAudit
S0004Question not draftedUnassigned——candidate—AnnotateAudit
S0005Question not draftedUnassigned——candidate—AnnotateAudit
S0006Question not draftedUnassigned——candidate—AnnotateAudit
S0007Question not draftedUnassigned——candidate—AnnotateAudit
S0008Question not draftedUnassigned——candidate—AnnotateAudit
S0009Question not draftedUnassigned——candidate—AnnotateAudit
S0010Question not draftedUnassigned——candidate—AnnotateAudit