Verified Global Leaderboard & 5-Spoke Model Arena
Verified Global Leaderboard & 5-Spoke Model Arena
The Vox Verified Global Leaderboard (VoxLeaderboardArena) presents cryptographically verifiable community stewardship rankings, Collegium guild standings, and 5-spoke MENS model evaluation benchmarks ((model_id, harness_id, config_digest) tuples with 95% Wilson confidence intervals). All global standings are computed exclusively by vox-ops-ludus from HMAC-verified GitHub webhooks and Ed25519-signed benchmark receipts, linked in an append-only BLAKE3 hash chain, and published as an Ed25519-signed VerifiedLeaderboardSnapshotV1 snapshot with 100% static offline fallback.
Verified Global Arena & 5-Spoke Benchmark Standings
Zero client score writes. Every entry is derived from HMAC-verified GitHub webhooks or Ed25519-signed 5-spoke evaluation receipts and chained with append-only BLAKE3 proofs.
| Rank | Steward | Tier | Lumens | PRs / Reviews | Bench Lifts | 7D Velocity | Cryptographic Proof |
|---|---|---|---|---|---|---|---|
| #1▲1 | autonomic-core | Archon of the Compiler | 2,480 | 18 / 34 | +6 | ||
| #2— | mens-curator | Praetor of MENS | 1,960 | 14 / 27 | +9 | ||
| #3▲2 | clavis-sentinel | Tribune of Trust | 1,540 | 11 / 29 | +3 |
| Rank | (Model, Harness, Config) Tuple | Pass@1 (95% Wilson CI) | Δ vs Base | 5-Spoke Breakdown (Vox/Rust/Fnd/Tool/Ctx) | Contamination & Integrity |
|---|---|---|---|---|---|
| #1 | mens/qwen3.5-27b:five-spoke-qlora-r2 | 100.0% | +76.0% | 100% · 100% · 100% · 100% · 100% | Held-Out CleanZero Test Tamper |
| #2 | mens/qwen3.5-27b:five-spoke-qlora-r1 | 76.0% | +52.0% | 80% · 80% · 80% · 80% · 60% | Held-Out CleanZero Test Tamper |
| #3 | qwen/qwen3.5-27b-base | 24.0% | 20% · 40% · 20% · 20% · 20% | Held-Out CleanZero Test Tamper |
| Rank | Collegium | Pontifex | Members | Total Lumens | Verified Contributions | Chain Anchor |
|---|---|---|---|---|---|---|
| #1 | Ordo Compiler Forgers | autonomic-core | 14 | 6,840 | 92 | a11f3d231163… |
| #2 | Schola MENS Architects | mens-curator | 11 | 5,120 | 74 | a11f3d231163… |
Append-Only BLAKE3 Ledger Tip
c94e8b7a6d5f4e3c2b1a0987654321fedcba0987654321fedcba0987654321fe
Ed25519 Snapshot Signer Public Key
ea4a6c63e29c520abef5507b132ec5f9954776aebebe7b92421eea691446d22c
AntiCheatPolicyV1 Rules Enforced
- Zero client-submitted score endpoints (server-ingested proofs only).
- Self-merged PRs without independent review award
0Lumens. - AST / token diff threshold requires ≥ 10 non-comment nodes.
- Daily diminishing returns (wn = 0.5max(0, n − 5)) cap farming loops.
Verification Architecture
- Zero Client Score Writes (
LEADERBOARD-01): Localvox-gamifySQLite tables track solo developer progression only (canonical_user_id()). Global standings are derived exclusively from server-verified upstream events (ludus_community.verified_contributions). - Append-Only
BLAKE3Proof Chain &Ed25519Snapshot Signing (LEADERBOARD-02): Every verified contribution recordsprev_row_chain_hash_hexandrow_chain_hash_hexanchored to a 64-zero genesis hash, signed withvox-ops-ludus’sEd25519key (VerifiedLeaderboardSnapshotV1). - 5-Spoke Model & Mens Arena (
LEADERBOARD-03): Ranks(model_id, harness_id, config_digest)tuples across Vox Codegen, Rust Maintenance, Foundational Syntax, Tool/MCP Calling, and Context Engineering with 95% Wilson confidence intervals and contamination guards.