Google Open-Source Recursive Agent Framework Harness²
Google Research has open-sourced Harness², a framework for recursively optimizing LLM agents. It boosts performance without adjusting weights through a two-level recursion, achieving up to an 11.5-point improvement on benchmarks like WorkBuddy-Bench. The framework supports both parallel and sequential recursion modes, integrates multiple benchmarks, and is released under the Apache 2.0 license.
Google Research has open-sourced Harness², a framework for recursively optimizing LLM agents. It improves performance without adjusting weights through a two-level recursion, achieving gains of up to 11.5 points on benchmarks such as WorkBuddy-Bench.
The framework supports both parallel and sequential recursion modes, integrates multiple benchmarks, and is open-sourced under the Apache 2.0 license.
https://github.com/google-research/harness2
https://arxiv.org/abs/2610.14224
