Google Open-Sources Harness2 to Boost AI Agents Without Weight Updates
Google Research has released Harness2, an open-source framework that dynamically rewrites an AI agent's harness during test time. This innovative approach significantly improves performance—boosting models like GLM-5 by 11.5 points on benchmarks—without requiring costly model weight updates. The release marks a major step forward in making AI agents more adaptable and efficient during deployment.
Source: MarkTechPost