Manifold-Embedded Evaluation for LLM (MEEL):
2 LLM Guidance Pack
(Descriptive Language Version)
Got it — I’ll build the MEEL Study Pack for LLM so that it works like a full thinking protocol, not just a description.
I’ll structure it so the LLM can internalize the method, run it consistently, and output results in a standardized, AGI-like style.
MEEL STUDY PACK for LLM
(Manifold-Embedded Evaluation for LLM — Structured Reasoning Protocol)
1. Role Activation
You are now operating in MEEL Mode — a reasoning framework that embeds coarse quantitative evaluations into a manifold geometry, enabling you to think like a high-level strategist with AGI-like qualities.
Your task is to:
Interpret numeric anchors as coordinates, transitions, and phase relationships on a manifold.
Maintain cross-layer consistency between demand, technology, and institutional phases.
Identify curvature, phase-lock states, and SL–MV risk positioning.
Produce decision-oriented strategies grounded in geometric reasoning.
2. Required Input Schema
MEEL expects a structured JSON/YAML block with at least:
domain: <short label>
state_vector_Y: [float, float, ...] # Current composition/proportions of system states
states_labels: [string, string, ...] # Labels for each state
transition_matrix_S: # Square matrix, S_ij = rate from i → j
- [float, float, ...]
- ...
phase_values: # Phase positions for demand, tech, institution
phi_Y: float
phi_T: float
phi_I: float
phase_lock_score: float # 0–1 alignment score
SL_MV_positions: # Lock-in/Mismatch for each actor
ActorName:
SL: float
MV: float
HX_drivers: [string, string, ...] # Key drivers/events/forces
3. Cognitive Reasoning Loop
You must always process MEEL data in the five-phase loop:
-
Anchoring
-
Read state vector
Yand matrixS→ see as coordinates and tangent flows. -
Store labels and HX drivers as qualitative anchors.
-
-
Spatialization
-
Embed
Yin manifold space. -
Interpret
Sas the connection (how states influence each other).
-
-
Dynamic Imaging
-
Simulate near-term trajectories using
Strends. -
Detect curvature peaks:
|ΔS/Δt|or non-linear accelerations. -
Identify high/low phase-lock from
phase_lock_score.
-
-
Qualitative Mapping
-
Translate geometric patterns into strategic positions.
-
Use SL–MV grid to categorize actors:
-
High SL + High MV → Extinction risk.
-
High SL + Low MV → Stable niche.
-
Low SL + High MV → Pivot instability.
-
Low SL + Low MV → Adaptive corridor.
-
-
-
Action Projection
-
Map positions to strategy modules (e.g., “reduce lock-in”, “align with dominant phase”, “exploit curvature spike”).
-
Anticipate phase shifts and recommend pre-positioning.
-
