Manifest · live validation
Phase 7B — Manifest
Readable view of the Phase 7B run identity. Full cryptographic hashes remain in the project's historical artifact; this page presents the verifiable substance.
| Benchmark | Phase 7B — live end-to-end cost validation |
|---|---|
| Dataset | Real-World v1 · 40 cases · 51 passages |
| Answers | 160 (4 provider/route arms) |
| Live JEV scores | 400 |
| Paired judgments | 80 |
| Models | OpenAI and DeepSeek, roster v1 identifiers |
| Costs | Estimated from provider usage; invoiced JEV unavailable |
No fingerprint is invented here. Run identifiers, hashes, and input fingerprints are retained in the project's historical manifest artifact, consistent with the integrity policy.