paper_type.json (275B)
1 { 2 "paper_type": "benchmark-creation", 3 "reason": "The paper's primary contribution is curating and introducing two new benchmark datasets (4,275 and 894 traces) for detecting silent failures in multi-agentic systems, with experiments validating the benchmark's utility." 4 }