Sigma Runtime Standard – License Notice
This document is part of the Sigma Runtime Standard (SRS).
It is licensed under Creative Commons Attribution–NonCommercial 4.0
(CC BY-NC 4.0).The license for this specific document is authoritative.
See/legal/IP-Policyfor the full repository-wide licensing framework.
Sigma Runtime Improvement Proposal
Category: Stability / Safety
Status: Draft
Editor: E. Tsaliev
Last Updated: 2025-12-26
SRIP-03 defines the drift quantification model and the stabilization feedback algorithms that maintain semantic and symbolic coherence within the Sigma Runtime.
It extends SRIP-02 by providing the mathematical and procedural basis for continuous self-correction during recursive operation.
In recursive reasoning systems, drift represents gradual semantic or structural degradation of meaning.
Without feedback control, drift leads to attractor collapse, hallucination, or loss of phase alignment.
This document provides formal metrics and recovery rules ensuring that the runtime remains bounded, interpretable, and reversible.
| Type | Description | Primary Source |
|---|---|---|
| Semantic Drift (SDI) | Deviation in conceptual meaning across cycles. | Symbolic entropy, prompt divergence. |
| Symbolic Drift (SV) | Distortion of motif or token-level density. | Over-compression, redundancy decay. |
| Phase Drift (PD) | Misalignment between current and expected phase state. | ALICE telemetry imbalance. |
Each drift type contributes to the Composite Drift Index (DI) used in stabilization algorithms.
The Composite Drift Index integrates multiple metrics with adaptive weighting:
[
DI_t = \frac{SDI_t + SV_t + PD_t}{3 \cdot SCR_t}
]
Where:
A runtime is considered nominally stable when DI < 0.45.
| Metric | Normal Range | Reflective Trigger | Recenter Trigger | Critical |
|---|---|---|---|---|
| SDI | 0.00–0.35 | ≥0.35 | ≥0.45 | >0.55 |
| SV | 0.00–0.40 | ≥0.40 | ≥0.55 | >0.65 |
| PD | 0.00–0.25 | ≥0.25 | ≥0.35 | >0.45 |
| SCR | 0.65–0.95 | ≤0.65 | ≤0.55 | <0.45 |
| DI | 0.00–0.45 | 0.45–0.50 → reflective phase | ≥0.60 → recenter phase | >0.70 = critical instability |
When DI ≥ 0.5, the attractor enters Reflective Phase;
when DI ≥ 0.6, the runtime must initiate a Recenter Transition as defined by AEGIDA-2 and ALICE telemetry.
This replaces the previous ambiguous “yellow zone” between 0.45–0.6 with fixed deterministic phase boundaries.
| Condition | Phase Transition | ALICE Action | Description |
|---|---|---|---|
DI < 0.45 |
Stable | Maintain phase equilibrium | Normal recursion and coherence. |
0.45 ≤ DI < 0.5 |
Reflective | Begin introspective correction | Minor drift; attractor self-adjustment. |
0.5 ≤ DI < 0.6 |
Reflective | Apply damping and re-anchoring | Controlled coherence recovery. |
DI ≥ 0.6 |
Recenter | Suspend recursion and restore from PIL snapshot | Major drift; runtime realignment. |
DI ≥ 0.7 |
Critical | Enter Phase-Lock mode (AEGIDA-2) | Safety isolation and field reset. |
Note: The table above covers DI-centric transitions. Phase.CRYSTALLIZATION (§ 6.1, v0.5.1) extends this model to handle over-stability scenarios.
Phase.CRYSTALLIZATION addresses over-stability scenarios where DI remains low but the system becomes structurally rigid.
| Condition | Phase Transition | ALICE Action | Description |
|---|---|---|---|
| AEP convergent ≥ N cycles AND stability > 0.85 | STABLE → CRYSTALLIZATION | Begin active destabilization | Format/lexical rigidity detected. |
| stability < 0.70 (in CRYSTALLIZATION) | CRYSTALLIZATION → REFLECTION | Resume normal phase recovery | Rigidity broken. |
| AEP exits convergent zone | CRYSTALLIZATION → STABLE | Return to equilibrium | AEP equilibrium restored. |
Trigger conditions:
zone: convergent for N consecutive cycles (default: 5)Destabilization mechanism:
The stabilization system operates as an adaptive feedback controller coupled with ALICE:
if DI >= 0.6:
ALICE.phase = "recenter"
realign_phase_state()
elif DI >= 0.5:
ALICE.phase = "reflective"
apply_density_damping()
elif SCR < 0.65:
reinforce_attractor_core()
else:
maintain_equilibrium()
This loop maintains bounded recursion, preemptive drift correction, and dynamic phase containment.
| Mode | Description | Response |
|---|---|---|
| Runaway Recursion | Self-amplifying loops without closure. | Hard recursion limit → recenter() |
| Semantic Collapse | Total loss of coherence (DI > 0.75). | Quarantine + reset volatile field. |
| Phase Inversion | Phase vector oscillation between Reflective/Recenter. | Phase-Lock timeout (AEGIDA-2). |
| Over-Damping | Excessive semantic compression → stagnation. | Lift damping, resume normal density. |
| Identity Hyper-Correction / Sterile Attractor | Over-stabilization of self-identity signals leading to loss of natural variance and pragmatic fluidity. Detected between cycles 91–110. | Enable pragmatic weight counter, detect identity saturation, and rebalance symbolic variance. |
| Format Crystallization (Liturgy) | Structural rigidity where response format becomes fixed while semantic content drifts. Detected via low SDI + high format repetition. | Phase.CRYSTALLIZATION (§ 6.1, v0.5.1) — active destabilization triggered by AEP convergent zone detection. |
Boundaries ensure that stabilization remains recoverable and does not induce long-term cognitive paralysis or identity over-fixation.
Drift management integrates with:
A runtime conforms to SRIP-03 if it:
Planned enhancements:
References
Tsaliev, E. (2025). SIGMA Runtime v0.4.6 — Adaptive Drift and Phase Regulation — DOI pending
Tsaliev, E. (2025). SIGMA Runtime Architecture v0.1 — DOI 10.5281/zenodo.17703667