TraceBench Agent Tool Design Benchmark: How Instructions Reshape Agent Trajectories
TraceBench is a simulation-based benchmark that measures how agent instructions and tool design change the behavior of LLM agents. By generating root-cause attribution tasks from interpretable physical dynamical systems with known ground truth, it isolates whether domain context, labeled examples, console-based tools, or submission format most changes an agent’s trajectory and accuracy. Its headline result: agents explore data through numerical console output far more than visualizations, and direct-answer submission consistently beats programmatic script generation. ...