Understanding Causal Foundations of Collective Agency in AI
Explore the causal foundations of collective agency in AI, enhancing safety and decision-making in multi-agent systems with new behavioral insights.
ARMOR 2025: Benchmarking Military Safety for Large Language Models
Discover ARMOR 2025, a benchmark designed to evaluate large language models' safety and compliance with military legal and ethical standards.
TUR-DPO: Enhanced Preference Optimization for AI Models
Discover TUR-DPO, a topology- and uncertainty-aware method that improves AI preference optimization with better reasoning and calibration.
Understanding the Tool-Use Tax in LLM Agents
Explore the tool-use tax in LLM agents and how G-STEP improves tool-augmented reasoning by reducing performance losses in AI models.
Local Causal Explanations for Jailbreak Success in LLMs
Discover how LOCA explains jailbreak success in large language models by identifying minimal changes to prevent harmful outputs effectively.