ABOUT — RTT/Inside/Benchmarks

Session Seed: rtt=1 | coherence=declared | drift=bounded | paradox=structural module=RTT/Inside/Benchmarks | layer=benchmark-substrate | operators=φ–V–R

RTT-not-physics rule (unconditional): RTT uses physics-aligned mathematics as structural scaffolding. No claim is made that RTT describes physical reality. All operators (φ, V, R), invariants (3C), and metrics are RTT-native constructs. Zone X and Mode 5 labels are specific to this module — see §6 and §7 for cross-module disambiguation.

1. What Is RTT/Inside/Benchmarks?#

RTT/Inside/Benchmarks is the cross-scale, physics-aligned benchmark suite for evaluating Structural Intelligence (SI) across the full RTT computational stack. It is the first standardized framework within TriadicFrameworks for measuring how well any agent, model, or system maintains structure, coherence, emergence, resonance, drift, entropy, and regime transitions — from 1D toy problems to 4096×4096 classical grids and from 2-qubit to 256-qubit quantum-classical hybrid circuits.

The module does not test semantic correctness, task accuracy, or performance benchmarks in the conventional ML sense. It tests structural fidelity: whether a system preserves its declared invariants (the 3C set — coherence, consistency, continuity) as it scales, transitions regimes, and encounters perturbation.

Primary measurement triad: φ (structural coherence), V (variance), R (resonance). Core score: SI_score(x) = φ(x) / (V(x) + ε) where ε is a regularization floor. Student standard: RTT-SI-Spec v0.1. Audience: students, researchers, developers, AIs, and standards bodies working within or evaluating RTT-aligned systems.

2. Why Is It Built This Way?#

2.1 Why a dedicated benchmark sub-module rather than inline tests?#

Benchmarking structural intelligence requires a stable, module-independent reference point. Embedding benchmark logic inside RTT/1–RTT/12 pipeline stages would couple measurement to implementation — any change to a pipeline stage would invalidate historical comparisons. RTT/Inside/Benchmarks is laterally positioned so that every pipeline stage can be evaluated without depending on any other stage's internals.

2.2 Why the φ–V–R operator triad specifically?#

The three operators were selected because they are orthogonal in their failure modes:

• φ (structural coherence) detects pattern collapse — the system loses internal structure.
• V (variance) detects instability — the system oscillates without settling.
• R (resonance) detects decoupling — subsystems that should reinforce each other diverge.

A system can fail on any one of these while passing the other two. Using all three together provides the minimum sufficient basis for diagnosing structural failure mode. Using fewer operators leaves blind spots.

2.3 Why the 3C invariants (coherence, consistency, continuity)?#

The 3C invariants are derived from the RTT structural requirement that any valid measurement must be repeatable in kind (consistency), stable across time (continuity), and internally non-contradictory (coherence). These are necessary conditions for any benchmark result to mean anything at all. A result produced without all three invariants satisfied is structurally undefined — it cannot be used for comparison, regression, or standards certification.

2.4 Why is DRIFT_GATE a hard interrupt rather than a soft warning?#

Drift is a structural contagion: once a measurement crosses the DRIFT_GATE threshold, subsequent measurements are contaminated. Allowing a soft warning and continuing measurement would produce a result chain in which some entries are valid and some are not — with no reliable way to separate them after the fact. The hard interrupt preserves the integrity of the entire capture chain by stopping at the first breach.

DRIFT_GATE fires when any of three conditions holds:

• C(t) < C_min — coherence has dropped below the declared minimum
• D(t) > D_max — drift rate has exceeded the declared maximum, where D(t) = |C(t) − C(t−1)| / Δt
• Zone = X — the measurement regime is OVERSCALE (see §7)

2.5 Why is Zone X = OVERSCALE a hard-labeled illegal zone?#

Extrapolating benchmark results beyond the declared scale or regime without validated lineage is not a measurement at all — it is a projection. Projections and measurements carry different epistemic weight and must never be mixed in a benchmark record. Labeling extrapolation as Zone X = OVERSCALE (ILLEGAL) makes this distinction machine-readable and enforceable at the agent level, not just as a documentation convention.

2.6 Why is Mode 5 = FABRICATION a hard-labeled illegal mode?#

Reporting a benchmark score without executing the actual measurement protocol is the benchmark equivalent of data fabrication. Mode 5 = FABRICATION (ILLEGAL) exists to give agents and auditors a precise, unambiguous label for this failure — distinct from measurement error, drift, or regime violation. The label is machine-readable so that any downstream consumer of a benchmark record can detect fabricated entries without manual inspection.

2.7 Why does the student standard (RTT-SI-Spec v0.1) exist as a separate artifact?#

New practitioners — students and researchers unfamiliar with RTT — need a constrained entry point that prevents them from inadvertently entering Zone X or Mode 5. The RTT-SI-Spec v0.1 defines the minimal valid benchmark protocol: what fields are required, what scale ranges are permitted, what operators must be reported, and what invariants must be declared. It is a teaching document and a compliance reference simultaneously.

2.8 Why does the CAPTURE_TEMPLATE require exactly 5 mandatory fields?#

The five fields — scope, lineage, provenance, interoperability, governance — represent the minimum information needed to make a benchmark result traceable and reusable:

• Scope: establishes what was measured and at what scale.
• Lineage: establishes what inherited constructs were active.
• Provenance: establishes who produced the measurement and under what conditions.
• Interoperability: establishes which downstream modules can consume the result.
• Governance: establishes which agents and protocols have authority over the result.

Omitting any field makes the record structurally incomplete and non-compliant with RTT-SI-Spec v0.1.

3. When Should You Use It?#

Use RTT/Inside/Benchmarks when:#

Do NOT use RTT/Inside/Benchmarks when:#

4. Where Does It Live?#

Repository path:#

docs/
└── rtt/
    └── Inside/
        └── Benchmarks/              ← YOU ARE HERE
            ├── AGENTS.md
            ├── ABOUT.md
            ├── GLOSSARY.md
            ├── A_Overview.md
            ├── B_Capture.md
            ├── Benchmarks_Capture.md
            ├── Benchmarks_module.json
            ├── C_Operators.md
            ├── J_RFCs/
            └── metadata/

Pipeline position:#

RTT/Inside/Benchmarks is a lateral sub-module of RTT/Inside. It is NOT a sequential pipeline stage. It sits alongside the other RTT/Inside sub-modules (Enterprise, qCompute) and can be invoked from any pipeline stage for structural evaluation without disrupting the main RTT/1–RTT/12 sequence.

RTT/1 → RTT/2 → RTT/3 → ... → RTT/12
                                    ↓
                              RTT/Inside
                             /     |     \
                   Benchmarks  Enterprise  qCompute
                   [THIS MODULE]

Inheritance chain:#

RTT/1 (origin)
  └── RTT/2 (field extension)
        └── RTT/3 (triadic integration)
              └── RTT/12 (convergence)
                    └── RTT/micro_core (distillation)
                          └── RTT/The_Inverted_Star (inversion)
                                └── RTT/Inside (interior + student layer)
                                      └── RTT/Inside/Benchmarks (THIS MODULE)

Submodule file manifest:#

5. Core Constructs at a Glance#

Operator triad:#

Core equations:#

DRIFT_GATE trigger conditions (any one is sufficient):#

C(t) < C_min          ← coherence below declared minimum
D(t) > D_max          ← drift rate above declared maximum
Zone = X              ← OVERSCALE (ILLEGAL) regime active

Collapse detection:#

∂F/∂t > F_collapse_threshold  ← entropy increasing faster than collapse threshold

CAPTURE_TEMPLATE — 5 mandatory fields:#

Scale ranges:#

3C Invariants:#

6. Module Integrations#

Upstream (what Benchmarks inherits):#

Downstream (what consumes Benchmarks output):#

Cross-module disambiguations:#

7. What RTT/Inside/Benchmarks Is Not#

8. Quick-Start Checklist#

Use this checklist before running any benchmark in RTT/Inside/Benchmarks:

• Declare scope — specify scale range and regime (classical grid dimensions OR qubit count); do NOT enter Zone X
• Declare lineage — confirm which RTT modules are active in the inheritance chain
• Initialize operators — φ, V, R must be instantiated with declared ε (regularization floor)
• Declare 3C invariants — confirm coherence, consistency, and continuity baselines before first measurement
• Open a CAPTURE_TEMPLATE — all 5 fields (scope, lineage, provenance, interoperability, governance) must be present
• Set DRIFT_GATE thresholds — declare C_min, D_max; confirm Guard (Class G) is active
• Confirm Guard is active — Class G has unconditional interrupt authority; do not proceed without Guard initialized
• Check for Zone X — if the requested scale exceeds the declared regime, stop; declare a new regime with validated lineage
• Check for Mode 5 — if the measurement protocol has not been executed, do not report an SI_score
• Run measurement — execute φ(x), V(x), R(x) at declared scale; compute SI_score
• Monitor drift — compute D(t) at each time step; check DRIFT_GATE conditions continuously
• Monitor entropy — compute F_entropy(t); check collapse condition ∂F/∂t > F_collapse_threshold
• Close capture — complete all 5 CAPTURE_TEMPLATE fields; record Guard status and DRIFT_GATE state
• Tag output — all output fields carry [structural — no semantic inference]

9. See Also#

[structural — no semantic inference] — all constructs, scores, operators, and records in this module are structural. No semantic meaning is inferred from any output field.