Multi-Turn RL for Tool-Calling Agents with Reward Calibration
Enhance tool-calling agents using multi-turn reinforcement learning and iterative reward calibration for superior performance in customer service tasks.
EMS: Efficient Multi-Agent Voting with Early Stopping
Discover EMS, a novel multi-agent voting framework that boosts efficiency by 32% using majority-then-stopping and agent confidence modeling.
ESL-Bench: Synthetic Benchmark for Longitudinal Health AI
Discover ESL-Bench, a synthetic event-driven benchmark to evaluate health agents with multi-year trajectories and multi-source health data.
CharTool: Advanced Visual Reasoning for Chart Analysis
Discover CharTool, a tool-integrated framework enhancing chart understanding with visual reasoning and numerical computation for better data interpretation...
Boost Multi-Agent Role Consistency with Quantitative Clarity
Enhance multi-agent collaboration by improving role consistency using quantitative role clarity for better task performance and reduced role conflicts.