Atlas · GenAI 2026
Explainable AI
Explainable AI (XAI) for GenAI
conceptPeak: 2025Explainability & FairnessAI consensus: 1/3
Prerequisites
Explaining Transformer decisions (attention visualization, probing, feature attribution) requires understanding the architecture
Recommended reference
Bills et al. (2023) 'Language Models Can Explain Neurons in Language Models' — OpenAI; mechanistic interpretability for Transformers
Notes from AI deep research
Anthropic Opus
Bills (2023) 'LMs Can Explain Neurons'. Mechanistic interpretability. W EU: prawo do wyjasnienia
Google Deep Think
Mechanistyczna interpretowalność [G#87]
Related skills
- → is subcategory of: AI Risk Management(1/3)
- → is subcategory of: AI Ethics(1/3)
- → is an instance of: Model Evaluation(1/3)
- → is an instance of: AI Fairness(1/3)
- ← is an instance of: Lime(0/3)
- ← is an instance of: SHAP(0/3)