@atlasauth/pca-control-evals 0.4.0 → 0.4.1
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/package.json +2 -2
package/package.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "@atlasauth/pca-control-evals",
|
|
3
|
-
"version": "0.4.
|
|
3
|
+
"version": "0.4.1",
|
|
4
4
|
"description": "Empirical adversarial control-evaluation harness for PCA: a seeded red-team/blue-team game (Redwood-style AI Control) that scores safety = P(a possibly-misaligned agent cannot cause an unacceptable outcome) + usefulness against PCA's verifier/budget/step-up/DLP/immune controls. Complements the analytic Σrisk ≤ bMax/κ bound.",
|
|
5
5
|
"keywords": [
|
|
6
6
|
"ai control",
|
|
@@ -32,7 +32,7 @@
|
|
|
32
32
|
"./package.json": "./package.json"
|
|
33
33
|
},
|
|
34
34
|
"dependencies": {
|
|
35
|
-
"@atlasauth/pca": "0.4.
|
|
35
|
+
"@atlasauth/pca": "0.4.1"
|
|
36
36
|
},
|
|
37
37
|
"devDependencies": {
|
|
38
38
|
"typescript": "^5.6.3",
|