Preventing Capability Loss in Self-Evolving LLM Agents
Explore how Capability-Preserving Evolution (CPE) helps self-evolving LLM agents maintain skills and prevent capability degradation during lifelong adaptat...
How Attention Heads Influence Persuasion in LLMs
Discover how a few attention heads reroute large language models, revealing key mechanisms behind LLM persuasion and AI safety.
Dynamic ESG Constraints for Smarter Portfolio Optimization
Discover how dynamic ESG constraints improve sequential portfolio optimization, balancing sustainability with financial performance in volatile markets.
Evaluating Strategy Diversity in LLM Math Reasoning
Discover how evaluating strategy diversity alongside accuracy enhances large language models' mathematical reasoning capabilities and flexibility.
PiCA: Pivot-Based Credit Assignment for Better RL Search Agents
Discover PiCA, a novel pivot-based credit assignment method boosting reinforcement learning in LLM search agents for complex tasks.