rigel-text-psychology

Rigel Text Psychology Expert

Part of the Rigel Constellation | Developed by Tenzro Labs

Model Description

Rigel Text Psychology Expert is the text analysis component of Rigel, a multi-modal AI system designed to understand human creativity and expression. Built on Google's Gemma-3-4b-it, this model specializes in psychological analysis of written content including letters, diaries, and biographical materials. As part of the broader Rigel ecosystem, it works alongside Rigel Vision, The Siren (audio analysis), and other specialized models to provide comprehensive multi-modal psychological insights.

Rigel Ecosystem Overview

Rigel explores how AI can help us better understand human creativity through multi-modal intelligence. Van Gogh serves as the initial proof-of-concept and pilot project, demonstrating the methodology before expanding to other creative figures. The "Becoming Vincent" pilot uses Van Gogh's extensive documentation (800+ letters, 2,100+ artworks) to validate systematic analysis of creative expression across multiple modalities.

The Rigel Model Constellation

Rigel Text (This Model): Psychological analysis of written content using clinical frameworks Rigel Vision: Visual art analysis for psychological indicators through brushwork, color, composition
The Siren (Rigel Audio): Speech and music analysis for emotional and psychological patterns Rigel Temporal: Time-series psychological tracking across creative periods Rigel Comparative: Cross-individual pattern recognition and psychological profile comparison Rigel Fusion: Meta-learning ensemble combining all modalities for comprehensive assessment

Van Gogh as Proof-of-Concept

Why Van Gogh for Initial Validation:

• Unprecedented Documentation: 800+ letters, 2,100+ artworks spanning entire artistic journey
• Multi-Modal Expression: Rich connections between written thoughts and visual creation
• Clear Periods: Distinct artistic phases showing psychological and creative evolution
• Scholarly Foundation: Decades of art historical research providing validation baseline
• Universal Research Interest: Subject that enables cross-disciplinary collaboration and validation

Pilot Project Validation: If Rigel can accurately analyze Van Gogh's documented creative patterns and psychological states, it establishes methodology for broader applications in digital humanities and creative analysis.

Institutional Foundation

Rigel is developed under Tenzro Labs, an AI research platform backed by Google for Startups with access to expertise from Google, Meta, Snap, LiveNation, Sony Music, and J.P. Morgan. The project integrates with PRVNZ (another Tenzro Labs initiative) for verifiable content provenance and operates on Tenzro's distributed AI computing infrastructure.

Model Details

• Base Model: unsloth/gemma-3-4b-it-unsloth-bnb-4bit
• Model Type: Causal Language Model with LoRA Fine-tuning
• Training Method: Unsloth + LoRA (Low-Rank Adaptation)
• Specialization: Clinical Psychology, CBT, Attachment Theory, Personality Psychology
• License: Gemma License
• Language: English

Training Infrastructure

• Hardware: NVIDIA A100-SXM4-40GB (40GB VRAM)
• Platform: Rigel Text Training System
• Framework: Unsloth 2025.8.2 + Transformers 4.55.0
• Precision: BFloat16 with 4-bit quantization
• Training Duration: 3 epochs, 3,300 steps

Training Data

The model was trained on a comprehensive psychology dataset including Van Gogh correspondence analysis as part of the proof-of-concept pilot project:

• Training Examples: 17,582 conversations
• Validation Examples: 1,954 conversations
• Total Dataset: 19,536 psychology examples
• Specialized Content:
  • Van Gogh letters and biographical analysis (pilot proof-of-concept)
  • Big Five Personality Framework examples
  • DSM-5 clinical psychology content
  • CBT and cognitive distortion frameworks
  • Attachment theory applications
  • Art therapy and creative expression analysis
  • Professional psychology standards and ethical boundaries

Note: Specific dataset composition and validation metrics available in accompanying research documentation.

Data Compliance (Documented in Compliance Documentation)

• Empathy Compliance: 84.4% (training), 82.5% (validation) - verified through automated empathy marker detection
• Ethical Disclaimers: 100% coverage - professional boundary compliance validation
• Safety Violations: 0 across all datasets - comprehensive safety content filtering
• Professional Boundaries: Maintained throughout - clinical standard adherence verification
• Privacy Protection: Complete PII anonymization - GDPR compliance protocols applied

Full compliance methodology and validation details available in accompanying 2025 Psychology AI Dataset Compliance Documentation.

Intended Use

Primary Applications

• Proof-of-Concept Creative Analysis: Text component for Van Gogh pilot project validation
• Historical Psychological Analysis: Analysis of letters, diaries, and biographical content
• Educational Psychology Content: Professional psychology framework explanations and applications
• Therapeutic Framework Analysis: CBT, attachment theory, and clinical assessment applications
• Research and Academic Applications: Digital humanities and computational psychology research
• Art Therapy Support: Professional development and educational applications

Pilot Project Integration (Van Gogh Proof-of-Concept)

• Standalone Text Analysis: Independent psychological analysis of Van Gogh correspondence
• Multi-Modal Validation: Establishes baseline for integration with future Rigel Vision analysis
• Temporal Pattern Recognition: Demonstrates psychological timeline tracking capabilities
• Methodology Validation: Proves systematic approach for future creative figure analysis
• Research Foundation: Establishes framework for "Becoming Frida," "Becoming Virginia," and other projects

Supported Psychology Domains

• Cognitive Behavioral Therapy (CBT)
• Attachment Theory
• Personality Psychology (Big Five)
• Clinical Assessment and Diagnosis
• Developmental Psychology
• Cognitive Distortions and Therapeutic Interventions

Usage

from unsloth import FastLanguageModel
import torch

# Load Rigel Text model
model, tokenizer = FastLanguageModel.from_pretrained(
    model_name="tenzro/rigel-text-psychology",
    max_seq_length=4096,
    dtype=torch.bfloat16,
    load_in_4bit=True,
)

# Enable inference mode
FastLanguageModel.for_inference(model)

# Example: Van Gogh letter analysis
prompt = """<bos><start_of_turn>user
Analyze this excerpt from Van Gogh's letter: "I feel a terrible need of—shall I say the word—religion. Then I go out at night and paint the stars."

Please provide a multi-framework psychological analysis.<end_of_turn>
<start_of_turn>model
"""

# Generate psychological analysis
inputs = tokenizer(prompt, return_tensors="pt")
outputs = model.generate(**inputs, max_new_tokens=512, temperature=0.8)
response = tokenizer.decode(outputs[0], skip_special_tokens=True)
print(response.split("<start_of_turn>model")[-1])

# Example: General CBT framework application
prompt = """<bos><start_of_turn>user
What is cognitive behavioral therapy and how does it help with anxiety?

Please provide a professional psychological perspective.<end_of_turn>
<start_of_turn>model
"""

inputs = tokenizer(prompt, return_tensors="pt")
outputs = model.generate(**inputs, max_new_tokens=512, temperature=0.8)
response = tokenizer.decode(outputs[0], skip_special_tokens=True)
print(response.split("<start_of_turn>model")[-1])

Multi-Modal Integration Example

# Example of how Rigel Text integrates with other Rigel models
# (Note: Other Rigel models released separately)

def analyze_creative_work(text_content, image_path=None, audio_path=None):
    """
    Comprehensive creative analysis using multiple Rigel models
    """
    results = {}
    
    # Text analysis (this model)
    text_analysis = rigel_text.analyze(text_content)
    results['psychological_text'] = text_analysis
    
    # Vision analysis (when available)
    if image_path:
        visual_analysis = rigel_vision.analyze(image_path)
        results['psychological_visual'] = visual_analysis
    
    # Audio analysis (when available) 
    if audio_path:
        audio_analysis = the_siren.analyze(audio_path)
        results['psychological_audio'] = audio_analysis
    
    # Multi-modal integration
    comprehensive_analysis = rigel_fusion.integrate(results)
    return comprehensive_analysis

Model Performance

Validated Quality Metrics (See Compliance Documentation)

• Professional Terminology Usage: Consistent evidence-based psychology frameworks
• Psychology Framework Application: Big Five, CBT, Attachment Theory, Clinical Assessment
• Empathy and Validation: 84.4% compliance in training, 82.5% in validation (documented methodology)
• Therapeutic Boundary Maintenance: 100% professional disclaimer compliance (verified)
• Safety Standards: Zero harmful content violations (comprehensive filtering applied)

Pilot Project Results (Van Gogh Proof-of-Concept)

• Text Analysis Performance: Demonstrates systematic psychological analysis of historical correspondence
• Framework Integration: Successfully applies multiple psychology frameworks to creative content
• Professional Standards: Maintains clinical boundaries while providing educational insights
• Research Validation: Establishes methodology baseline for multi-modal creative analysis

Detailed performance metrics, validation methodology, and expert review processes documented in accompanying research materials.

Limitations

Professional Boundaries

• This model provides educational and informational content only
• Not a substitute for professional mental health care
• Cannot provide clinical diagnoses or treatment recommendations
• Should not be used for crisis intervention or emergency situations

Technical Limitations

• Based on training data knowledge cutoff
• May not reflect the most recent research developments
• Responses should be verified against current professional standards
• Performance may vary across different cultural contexts

Ethical Considerations

• Always includes professional disclaimers
• Encourages seeking qualified mental health professionals
• Maintains appropriate therapeutic boundaries
• Designed for educational rather than therapeutic use

Bias and Fairness

The model has been specifically trained to:

• Use inclusive and gender-neutral language
• Avoid demographic stereotypes
• Respect cultural diversity in psychological approaches
• Maintain professional objectivity

Bias mitigation techniques applied during training include demographic representation analysis and inclusive language standards.

Training Details

LoRA Configuration

• Rank (r): 16
• Alpha: 32
• Target Modules: q_proj, k_proj, v_proj, o_proj, gate_proj, up_proj, down_proj
• Dropout: 0.0 (Unsloth optimized)

Training Hyperparameters

• Learning Rate: 2e-4
• Batch Size: 2 per device, 8 gradient accumulation steps
• Optimizer: AdamW 8-bit
• Scheduler: Cosine with 0.1 warmup ratio
• Max Sequence Length: 4096 tokens
• Training Epochs: 3

Infrastructure Specifications

• GPU: NVIDIA A100-SXM4-40GB
• Memory Usage: ~39.4GB available
• CUDA Version: 12.4
• Compute Capability: 8.0
• Training Framework: Unsloth 2025.8.2

Framework Versions

• PEFT: 0.17.0
• TRL: 0.21.0
• Transformers: 4.55.0
• PyTorch: 2.6.0+cu124
• Datasets: 3.6.0
• Tokenizers: 0.21.4

Ethical Guidelines

Professional Standards Compliance

• American Psychological Association (APA) Ethics Code
• Clinical practice guidelines and boundaries
• Evidence-based psychology frameworks
• Mental health AI safety standards

Safety Measures

• Comprehensive content filtering
• Crisis resource guidance protocols
• Professional disclaimer requirements
• Harmful content prevention

Citation

If you use this model in your research or applications, please cite:

@model{rigeltext2025psychology,
  title={Rigel Text Psychology Expert: Van Gogh Proof-of-Concept for Multi-Modal Creative Analysis},
  author={Hilal Agil},
  organization={Tenzro Labs},
  project={Rigel - Understanding Human Creativity Through Multi-Modal Intelligence},
  note={Van Gogh pilot project demonstrating systematic psychological analysis methodology},
  year={2025},
  publisher={Hugging Face},
  url={https://huggingface.co/tenzro/rigel-text-psychology}
}

@project{rigel2025vangogh,
  title={Van Gogh Proof-of-Concept: Validating Multi-Modal AI Analysis of Creative Expression},
  author={Hilal Agil and Tenzro Labs},
  year={2025},
  note={Initial pilot project demonstrating systematic analysis methodology using Van Gogh's documented correspondence and creative timeline}
}

Cite TRL as:

@misc{vonwerra2022trl,
	title        = {{TRL: Transformer Reinforcement Learning}},
	author       = {Leandro von Werra and Younes Belkada and Lewis Tunstall and Edward Beeching and Tristan Thrush and Nathan Lambert and Shengyi Huang and Kashif Rasul and Quentin Gallou{\'e}dec},
	year         = 2020,
	journal      = {GitHub repository},
	publisher    = {GitHub},
	howpublished = {\url{https://github.com/huggingface/trl}}
}

License

This model is released under the Gemma License. Please ensure compliance with Google's Gemma terms of use.

Contact

Author: Hilal Agil
Organization: Tenzro Labs
Email: [email protected]
Project: Rigel - Understanding Human Creativity Through Multi-Modal Intelligence

For questions about this model, the broader Rigel ecosystem, or collaboration opportunities in computational creativity research, please contact the author. For clinical guidance, always consult qualified mental health professionals.

Related Projects:

• Van Gogh Proof-of-Concept: Initial pilot validating multi-modal creative analysis methodology
• Rigel Vision: Visual art psychological analysis (in development for future integration)
• The Siren (Rigel Audio): Music and speech psychological analysis (advanced development stage)
• PRVNZ: Verifiable infrastructure for digital content provenance
• Future Projects: "Becoming Frida," "Becoming Virginia" following successful Van Gogh validation

---

Disclaimer: This AI model is designed for educational, research, and creative analysis purposes. It is not intended to replace professional mental health care, diagnosis, or treatment. Always consult qualified mental health professionals for clinical concerns.