AI alignment
New research proposes a cognitive architecture framework to address the 'black box' problem in AI systems, focusing on transparency, alignment, and interpretability through structured reasoning pathways.
AI Safety
New research introduces Factor(T,U) framework that uses task decomposition to improve monitoring of untrusted AI systems, addressing critical safety challenges as models become more capable.
AI alignment
New research introduces MoralReason, a reasoning-level reinforcement learning approach that aligns LLM agents with moral decision-making frameworks. The method generalizes across diverse ethical scenarios using structured reasoning processes.
AI Safety
Researchers introduce CTRL-ALT-DECEIT, a novel benchmark for evaluating whether AI systems conducting automated R&D can engage in sabotage. The framework tests adversarial behaviors in AI agents with specific technical metrics.
AI Safety
Researchers propose fundamental shift from post-hoc alignment to intrinsic identity-based AI development, arguing current training methods create misaligned systems that require extensive correction after the fact.