#causal-analysis News & Analysis

3 articles tagged with #causal-analysis. AI-curated summaries with sentiment analysis and key takeaways from 50+ sources.

3 articles

AINeutralarXiv – CS AI · 3d ago7/10

🧠

Why Do Large Language Models Generate Harmful Content?

Researchers used causal mediation analysis to identify why large language models generate harmful content, discovering that harmful outputs originate in later model layers primarily through MLP blocks rather than attention mechanisms. Early layers develop contextual understanding of harmfulness that propagates through the network to sparse neurons in final layers that act as gating mechanisms for harmful generation.

AINeutralarXiv – CS AI · 3d ago7/10

🧠

Thought Branches: Interpreting LLM Reasoning Requires Resampling

Researchers demonstrate that interpreting large language model reasoning requires analyzing distributions of possible reasoning chains rather than single examples. By resampling text after specific points, they show that stated reasons often don't causally drive model decisions, off-policy interventions are unstable, and hidden contextual hints exert cumulative influence even when explicitly removed.

AINeutralarXiv – CS AI · Mar 34/107

🧠

Econometric vs. Causal Structure-Learning for Time-Series Policy Decisions: Evidence from the UK COVID-19 Policies

A research study compares econometric methods versus causal machine learning algorithms for analyzing time-series data to inform policy decisions, using UK COVID-19 policies as a case study. The research evaluates four econometric methods against eleven causal ML algorithms, finding that econometric methods provide clearer temporal structure rules while causal ML algorithms explore broader graph structures to capture more causal relationships.