M-3ToolEval

The 'M-3ToolEval' is a benchmark dataset used to evaluate the reliability of code-mode tool use in models by assessing their performance on tasks involving inter-tool contract compliance and execution feedback.

Papers using M-3ToolEval (2)

Self-Challenging Language Model Agents2025 · 1 citesRubricRefine: Improving Tool-Use Agent Reliability with Training-Free Pre-Execution Refinement2026