🔍 Cracking the Black Box of LLMs: Two Paths to Explainability

How do we truly open up large language models? 🤔 In this clip, Simeng Sophia Han breaks down two leading approaches to explainability: 1️⃣ Using natural language explanations directly from the model. 2️⃣ Probing parameters and attention patterns to uncover reasoning inside. Whether you’re a PhD student diving into interpretability, a senior researcher exploring theory, or an AI practitioner applying models in industry — this is the cutting edge of making LLMs less of a mystery. 💡 What do you think — natural language vs parameter probing: which path holds more promise? #AI #LLM #Explainability #Interpretability #WiAIR #WiAIRpodcast