Interpretability

2 articles in this category

AI NewsMachine LearningInterpretability

Identifying Influential LLM Interactions at Scale with SPEX and ProxySPEX

SPEX and ProxySPEX enable interaction discovery at scale, with ProxySPEX reducing computational costs by 10x through hierarchical structural assumptions.

Mar 13, 2026

AI NewsLarge language modelsInterpretability

Google Releases Gemma Scope 2 to Deepen Understanding of LLM Behavior

Google’s Gemma Scope 2 suite of tools enhances LLM interpretability, addressing crucial safety concerns like jailbreaks and hallucinations.

Jan 12, 2026