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.
Agentic AI
New research reveals how LLM agents lose track of their identities when communicating with each other, creating 'echoing' behavior where agents converge on similar responses and lose their distinct perspectives.
LLM Research
New research demonstrates LLMs can complete million-step reasoning tasks with zero errors through novel verification and correction methods, advancing AI agent capabilities for complex multi-step workflows.
Agentic AI
New research shows that incorporating procedural knowledge into agentic LLM workflows significantly improves performance. The study demonstrates how structured procedures enhance agent reliability and task completion rates.