Guide Labs

GuideLabs.ai develops interpretable foundation models in artificial intelligence, including Language Learning Models (LLMs), Diffusion models, and large-scale classifiers. This technology makes AI more understandable and controllable by humans, offering features such as reliable prompt attribution, concept explanation, and identification of influential data points.

Key features:
- Explainability and Steering: Models explain reasoning in human-understandable terms and can be controlled based on high-level concepts.
- Reliable Prompt Attribution: Identifies which parts of a prompt are most influential in the model's output.
- Data Attribution: Identifies tokens in pre-training and fine-tuning data that most influence the model's output.
- Concept Explanation: Allows customization with high-level, human-understandable features provided by domain experts.
- Multi-Modal Capabilities: Models can be trained or fine-tuned on any data modality.

How it works:
1. Users access the API to interact with interpretable models.
2. Users can fine-tune models with their own data.
3. High-level concepts can be inserted to steer model output.
4. Users generate responses with explanations for prompt attribution, concept importance, and influential data points.
5. Interaction is primarily facilitated through a web app.

Use of AI:
GuideLabs.ai's technology is built on interpretable foundation models, including LLMs, Diffusion models, and large-scale classifiers. These models are developed based on insights from machine learning literature.

Target users:
- Developers
- Data scientists
- Researchers interested in AI interpretability

How to access:
GuideLabs.ai operates under a proprietary model and is currently working with selected users to test these models. Early access is available through a waitlist.