Search papers, labs, and topics across Lattice.
This paper introduces DarwinX, an innovative approach to evolving the harness of an LLM agent through a population-based selection mechanism while keeping the model weights frozen. By implementing a preserve-and-extend contract and utilizing an archive for recombination, DarwinX enables self-improvement loops that enhance general agent competence across multiple benchmarks without regressing on other tasks. The results show significant performance improvements, with average gains of about 17 points across four benchmarks, demonstrating that harness evolution can lead to durable capabilities in LLM agents.
Evolving the harness of LLM agents can yield a 17-point performance boost without compromising their ability to generalize across tasks.
An LLM agent's capability depends not only on model weights but on its harness: prompts, tools, skills, and control flow. Self-improvement loops already edit harnesses, yet single-lineage search is path-dependent and local wins often regress other tasks. We introduce DarwinX, which treats self-evolution as selection over a population of harnesses with the model frozen: a preserve-and-extend contract admits only variants that extend coverage without regressing, an archive keeps alternative lineages for recombination, and failure-, teacher-, and self-derived evidence share one edit interface. Fitness comes from each benchmark's own verifier: no gold solutions, no hand-picked winners. Across four benchmarks that progressively separate the evolution signal from the test, one loop adds about 17 points on average: Terminal-Bench 2.1 rises +7.7 to 83.2% on a matched base and to the verified frontier at 84.7% on a stronger one; TerminalWorld's held-out split reaches 68.3%, ahead of every off-the-shelf agent; WebArena-Infinity real-task pass@1 rises from 43.5% to 93.0% audit-clean; and a Terminal-Bench 2.1 harness transfers unchanged to SWE-bench Verified. What evolves is general agent competence, not benchmark-specific patches, so it survives changes of task, verifier, and base model. A frozen model need not be a fixed agent: harness selection turns evaluation compute into durable capability.
