Crovia Bot commited on
Commit ·
0c38493
1
Parent(s): 3c20761
docs: define DSSE 1M as primary Crovia evidence
Browse files
README.md
CHANGED
|
@@ -1,33 +1,105 @@
|
|
| 1 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 2 |
|
| 3 |
-
|
| 4 |
|
| 5 |
-
|
| 6 |
-
|
| 7 |
-
|
| 8 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 9 |
|
| 10 |
## What this is
|
| 11 |
-
This is **not** a benchmark.
|
| 12 |
-
This is **not** a model.
|
| 13 |
-
This is an **audit artifact**.
|
| 14 |
|
| 15 |
-
|
| 16 |
-
|
| 17 |
-
|
| 18 |
-
|
| 19 |
-
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 20 |
|
| 21 |
-
|
| 22 |
-
|
| 23 |
-
|
| 24 |
|
| 25 |
-
|
| 26 |
-
- Raw datasets
|
| 27 |
-
- Proprietary processing logic
|
| 28 |
-
- Model internals
|
| 29 |
|
| 30 |
-
|
| 31 |
|
| 32 |
-
|
| 33 |
-
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
tags:
|
| 4 |
+
- crovia
|
| 5 |
+
- trust
|
| 6 |
+
- provenance
|
| 7 |
+
- auditability
|
| 8 |
+
- dsse
|
| 9 |
+
- evidence
|
| 10 |
+
pretty_name: "DSSE 1M — Verifiable Trust Drift Proof"
|
| 11 |
+
---
|
| 12 |
|
| 13 |
+
# DSSE 1M — Verifiable Trust Drift Proof
|
| 14 |
|
| 15 |
+
This repository is the **primary public evidence** of the Crovia project.
|
| 16 |
+
|
| 17 |
+
It demonstrates that a **measurable trust drift signal** can be observed:
|
| 18 |
+
|
| 19 |
+
- between two temporal states of a dataset
|
| 20 |
+
- with **no declared dataset change**
|
| 21 |
+
- using a reproducible, auditable pipeline
|
| 22 |
+
- on commodity hardware
|
| 23 |
+
- at the scale of **1 million records**
|
| 24 |
+
|
| 25 |
+
No interpretation is provided.
|
| 26 |
+
No attribution is claimed.
|
| 27 |
+
This repository contains **evidence only**.
|
| 28 |
+
|
| 29 |
+
---
|
| 30 |
|
| 31 |
## What this is
|
|
|
|
|
|
|
|
|
|
| 32 |
|
| 33 |
+
This dataset is a **cryptographically anchored evidence package** produced by Crovia.
|
| 34 |
+
|
| 35 |
+
It shows that:
|
| 36 |
+
|
| 37 |
+
- trust-related structural properties of a dataset
|
| 38 |
+
- can drift over time
|
| 39 |
+
- even when no official change is declared
|
| 40 |
+
- and that such drift can be **measured, reproduced, and verified**
|
| 41 |
+
|
| 42 |
+
The output is a signal — not a conclusion.
|
| 43 |
+
|
| 44 |
+
---
|
| 45 |
+
|
| 46 |
+
## What this is NOT
|
| 47 |
+
|
| 48 |
+
- ❌ Not a benchmark
|
| 49 |
+
- ❌ Not a model evaluation
|
| 50 |
+
- ❌ Not a claim against any dataset or organization
|
| 51 |
+
- ❌ Not a legal or ethical judgment
|
| 52 |
+
|
| 53 |
+
Crovia does **not** interpret the signal.
|
| 54 |
+
Crovia only makes it verifiable.
|
| 55 |
+
|
| 56 |
+
---
|
| 57 |
+
|
| 58 |
+
## Contents (root)
|
| 59 |
+
|
| 60 |
+
- `START_HERE.md`
|
| 61 |
+
→ minimal orientation for verification
|
| 62 |
+
|
| 63 |
+
- `EVIDENCE.ndjson`
|
| 64 |
+
→ full machine-readable evidence trace
|
| 65 |
+
|
| 66 |
+
- `HASHES.txt`
|
| 67 |
+
→ cryptographic anchors
|
| 68 |
+
|
| 69 |
+
- `SNAPSHOT_LATEST.json`
|
| 70 |
+
→ latest observed state (with evidence attached)
|
| 71 |
+
|
| 72 |
+
- `synth_1M_drift.ndjson`
|
| 73 |
+
→ aggregated drift signal (1M scale)
|
| 74 |
+
|
| 75 |
+
---
|
| 76 |
+
|
| 77 |
+
## Reproducibility
|
| 78 |
+
|
| 79 |
+
All artifacts in this repository are:
|
| 80 |
+
|
| 81 |
+
- deterministic
|
| 82 |
+
- hash-linked
|
| 83 |
+
- reproducible from inputs
|
| 84 |
+
- inspectable without proprietary tooling
|
| 85 |
+
|
| 86 |
+
If the signal is wrong, it can be disproven.
|
| 87 |
+
That is the point.
|
| 88 |
+
|
| 89 |
+
---
|
| 90 |
+
|
| 91 |
+
## Position in the Crovia ecosystem
|
| 92 |
+
|
| 93 |
+
This repository is the **reference proof**.
|
| 94 |
|
| 95 |
+
Other Crovia datasets (LAION examples, CEP capsules, conceptual datasets)
|
| 96 |
+
exist to illustrate, explore, or contextualize the system —
|
| 97 |
+
**not** to replace this evidence.
|
| 98 |
|
| 99 |
+
---
|
|
|
|
|
|
|
|
|
|
| 100 |
|
| 101 |
+
## One rule
|
| 102 |
|
| 103 |
+
> If a system affects trust,
|
| 104 |
+
> there must be evidence.
|
| 105 |
+
> And that evidence must be auditable.
|