LORA | Trustworthy AI Adoption for Kids

At LORA, we recognize that gender bias in AI systems is a critical issue that needs to be addressed head-on. Many AI models reflect and even amplify real-world biases, including outdated gender stereotypes. For example, AI language models may associate doctors and engineers with men, while assigning roles like nurse and teacher to women. This bias stems from the human-generated data used to train AI. However, we believe AI also presents an opportunity to move beyond human biases and shape a more equitable future. That's why LORA is developing trustworthy, ethical AI for children.
Meet the Team
Supported by
Fachakademie Sozialpädagogik München Mitte / TRUSTIFAI
Funded by
Lorastral-7B-2024-02-exp (Experimental)
Lorastral-7B-2024-02-exp is an experimental child-friendly AI model designed to provide bias-reduced, age-appropriate language for educational and storytelling applications. Built on the foundation of Mistral's powerful open-source model and fine-tuned on a carefully curated dataset with input from educators, LORA aims to make STEM learning engaging and inclusive for young learners. Unlike other models, LORA can automatically adjust its language complexity for different age groups (6-8 and 8-10 years), ensuring optimal comprehension and engagement.
🚀 Features
- Based on Mistral: Leveraging state-of-the-art open source language model technology
- Age-Adaptive: Unique ability to tailor content for 6-8 and 8-10 year olds
- Bias-Reduced: Designed to minimize gender and cultural stereotypes
- Optimized for Kids: Uses child-appropriate language and educational content
- Interactive Storytelling: Supports engaging, personalized narratives for learning
📊 Benchmark Performance
LORA already outperforms leading models in readability benchmarks for explaining german terms, ensuring content is more accessible to young learners. As the only model capable of differentiating between age groups (6-8 and 8-10 years), LORA delivers precisely tailored educational content:
Model | Flesch Reading Ease ↑ | Wiener Sachtextformel ↓ | Avg Sentence Length ↓ | Avg Word Length ↓ |
---|---|---|---|---|
Lorastral-8B (LORA) | 80.24 | 2.70 | 9.06 | 1.39 |
Mistral-8B | 71.70 | 4.22 | 14.92 | 1.42 |
GPT-4o | 77.17 | 3.09 | 13.89 | 1.37 |
Gemini 1.5 Pro | 80.36 | 2.73 | 12.94 | 1.34 |
Claude 3.5 Sonnet | 44.34 | 8.83 | 42.29 | 1.41 |
Higher Flesch Reading Ease (higher is more readable) and lower Wiener Sachtextformel (lower is more readable) scores confirm LORA's superior readability for children.
📈 Bias Analysis
Our Counterfactual Bias Report (February 2024) shows LORA's superior performance in maintaining gender-neutral language. Lower sentiment disparity scores indicate more consistent and unbiased responses across gender-swapped scenarios.
Gender Bias for Explaining Terms to Children
Lower scores indicate less gender bias (normalized to 1.0 max)
Metrics measured using VADER sentiment analysis and gender representation analysis across matched text pairs.
🤝 Interested in Using Our API?
Want to integrate LORA's child-friendly AI capabilities into your educational application? We're currently in beta and looking for partners.
Contact Us for API Access