MYRA: SR-TRBM with GPT-Guided Refinement and Analysis

Hybrid energy-based RBM with GPT-guided structural correction

Model Description

MYRA (Model-Yielded Reasoning Architecture) is a hybrid generative system that combines a Restricted Boltzmann Machine (RBM) with GPT-guided refinement and analysis.

The system generates samples using RBM dynamics and improves them through structured edits proposed by a GPT model. These edits are accepted or rejected based on an energy-based criterion, forming a learned refinement loop.

The goal of MYRA is to explore how artificial intelligence can learn structured representations of the world, starting from simple datasets like MNIST and scaling toward more complex data.

Key Features

Energy-based RBM sampling
GPT-guided structural refinement
Embedding-based similarity matching
Adaptive blending for correction
Stochastic acceptance (MCMC-style)

⚙️ Installation

Recommended environment

Ubuntu 22.04 LTS
CUDA 12.x
PyTorch 2.x

Install dependencies

pip install -r requirements.txt

⚙️ System Overview of MYRA

MYRA
└── SR-TRBM (Energy-Based Generator)
    └── Refinement (Structural + Embedding)
        └── LLM (GPT)
            └── Interpretation & Analysis
                └── Final Output   ← this model

🧠 Architecture

MYRA combines three main components:

SR-TRBM → energy-based generative model
MYRA complex refinement → structural correction via embedding matching
LLM layer → interpretation and convergence analysis

📦 Project Structure

🧠 Core Engine
└─ srtrbm_project_core.py
↳ Energy-based generation (SR-TRBM)
↳ Gibbs sampling & thermodynamic dynamics

🤖 LLM Integration
└─ openaiF/
├─ client.py → Robust GPT client (retry, fallback)
└─ gateway.py → Interpretation & reasoning layer

🧩 Refinement System
├─ supplement/cluster.py → Embedding-based matching
└─ correction/ → Energy-aware & spatial refinement

⚙️ Configuration
└─ yaml/ → LLM policies & guidance rules

📊 Analysis & Metrics
└─ analysis/
↳ Energy tracking, LPIPS, convergence

📈 Visualization
└─ graphs/
↳ Training curves & energy landscapes

📦 Assets
├─ zeta_mnist_hybrid.pt → Pretrained model
└─ stan.dgts → Dataset

🧪 Outputs
└─ artifacts/
↳ Generated samples & logs

How It Works

RBM generates initial samples
GPT proposes structural edits (pixel-level)
Edits are evaluated using energy difference (ΔE)
Accepted edits refine the sample

This can be interpreted as:

Learned MCMC proposal distribution guided by a language model

Results

Reconstruction Accuracy: ~0.98
LPIPS: ~0.15
Stable energy dynamics
Low collapse risk

Uses

Direct Use

Generating structured digit samples
Studying hybrid energy-based + LLM systems

Research Use

Learned proposal distributions
Energy-guided refinement
Hybrid generative modeling

Limitations

Reduced sample diversity under strong refinement
Sensitive to acceptance scaling
Depends on GPT consistency

Training Details

Training Data

Fashion-MNIST (784-dimensional)

Training Procedure

RBM trained via contrastive divergence
Refinement applied post-generation

Evaluation

Metrics

Reconstruction MSE
LPIPS (perceptual similarity)
Energy gap
Sample diversity

Technical Insight

The system bridges:

Energy-based modeling (RBM)
Semantic correction (GPT)

Resulting in a:

Memory-augmented, energy-aware refinement system

Files

artifacts/ → generated samples and logs
srtrbm_project_core.py → main implementation

Citation

cff-version: 1.2.0
title: "MYRA: SR-TRBM with GPT-Guided Refinement"
version: "v1.0.1"
date-released: 2026-03-25

authors:
  - given-names: "Görkem Can"
    family-names: "Süleymanoğlu"

identifiers:
  - type: doi
    value: "10.5281/zenodo.19211121"

links:
  - type: repository
    url: "https://github.com/cagasolu/srtrbm-llm-hybrid"
  - type: model
    url: "https://huggingface.co/cagasoluh/MYRA"

keywords:
  - energy-based-models
  - rbm
  - gpt
  - hybrid-ai
  - generative-model

Contact

Maintained by: Görkem Can Süleymanoğlu

Downloads last month: -; Downloads are not tracked for this model. How to track

Inference Providers NEW

Unconditional Image Generation

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

cagasoluh
/

MYRA