Tag
This paper presents a retrospective analysis of the CODS 2025 AssetOpsBench challenge, evaluating multi-agent AI systems on industrial tasks. It highlights discrepancies between public and hidden leaderboards and offers diagnostics for future agentic benchmarks.