Nvidia Research: The Harness, Not the Model, Is Key for AI Agents

Summary

New Nvidia research highlights that for long-horizon tasks, the software harness—the tools, memory management, and rules around a model—matters more than the model itself. Using a custom harness with improved memory handling and a supervisor component, researchers achieved a 100% score on the interactive reasoning benchmark ARC-AGI-3 with Claude Opus 5, compared to 30% without the harness, which was still the best among models tested. Nvidia product VP Adel El Hallak notes that an agent encompasses the model, scaffolding, runtime, and skills. Long-horizon tasks require many sequential decisions and are prone to errors; Microsoft's April research found all tested LLMs filled documents with errors on such tasks.