V5 Experimental Model
V5 is an archived experimental comparison artifact retained beneath the live V7 release. It remains separate from the live headline score while preserving its sidecars, transition-adjusted layer, realized-risk proxy, and validation comparison for auditability.
This report is results-oriented: it shows what the integrated V5 candidate currently does. The planning rationale and workstream framing stay on the V5 roadmap.
Status
Experimental only
not live in occupations
Structural validation
2/2
BLS and family checks versus V7
Realized validation
2/4
scorable short-run checks retained in the audit trail
Task-mode blended
492
occupations with mode-aware exposure blend
Median fragility
0.1185
concentration-weighted task fragility
What Is Integrated
Structural layer
Latent posterior + task modeSource-calibrated latent exposure stays the anchor, while task-mode proxies derived from weighted O*NET tasks adjust the exposure layer where the task evidence is strong enough.
Concentration layer
Fragility + reallocationHampole-style concentration now enters as demand fragility and reallocation capacity. A few highly exposed core duties are treated differently from diffuse exposure spread across many smaller tasks.
Adaptation layer
Augmentation + hybrid mobilityHeterogeneous augmentation and hybrid mobility become a bounded adaptation buffer, creating a transition-adjusted risk view rather than directly overwriting structural risk.
Realized-risk layer
Short runThe realized-risk proxy remains conservative and separate, reflecting archetype-specific short-run caps, adoption timing, postings resistance, and offset room rather than raw structural overlap.
Governance
Experimental onlyThis candidate is auditable because the sidecars remain published separately. Any future promotion still needs an explicit release decision.
Validation Snapshot
Structural checks
BLS rho
live -0.1387 → experimental -0.2007
Family rho
live -0.249 → experimental -0.4276
Realized-risk checks
Vacancy trend rho
-1
Hiring-net rho
0.5
Scorable, but still only a coarse 3-cluster diagnostic
Retrenchment rho
0
Current cluster series is too flat to score honestly
Employer-pressure rho
0.4871
positive coherence across archetypes, sample 435
Postings-support rho
0
negative coherence against direct occupation-level postings support, sample 93
Profile Mix
leveraged
99
occupations
shielded
189
occupations
retooling
149
occupations
balanced
125
occupations
Largest Changes
Structural risers
- Software developer (0.2799)
- Web and mobile applications developer (0.2147)
- Salesperson (door-to-door) (0.2117)
- Storekeeper (0.2069)
- Stock records clerk (0.2031)
Transition fallers
- Personnel/Human resource clerk (-0.1887)
- Ledger and accounts clerk (-0.1694)
- Crewing executive (ship) (-0.1568)
- Payroll/Wages clerk (-0.1449)
- Postal service counter clerk (-0.1439)
Highest realized risk
- Bookkeeper (0.3293)
- Pawnbroker (0.3188)
- Secretary (0.2876)
- Securities clerk (0.2731)
- Office clerk (including filing and copying) (0.2699)
Downloads