Vercy · benchmark directory
Benchmarks.
Each study keeps its instrument, its unedited raw runs and its result file in its own directory. This page lists those directories. The reports, with the findings and the reviews that invalidated earlier versions, are on the research index.
Tools enabled v1
Does giving the agent a shell and putting the data on disk change the balance between context representations?
Collaborative memory v1
When six teams write into one shared memory, does it matter whether ownership, the conflict rule and the release list live inside the record or in a governance layer beside it?
Personal dimension v1
For one person's own records, which part of a structured representation carries which capability?
Semantic grounding v1
Does a versioned definition layer make an agent more accurate when two organisations define the same words differently?
Memory retrieval v1
Does retrieval over a structured dimension beat flat notes on recall of a project's own history?
Index scale v1
Not a model study. A deterministic check that the reference implementation builds and queries an index over 5,000 objects, 5,000 bitemporal facts and 4,999 relations.