Search papers, labs, and topics across Lattice.
This paper systematically evaluates 30 budget-matched discovery harnesses across 12 model-problem pairs using over 3.1 million LLM rollouts to analyze the effectiveness of various design choices in autonomous discovery systems like OpenEvolve and TTT-Discover. The findings reveal that no single harness consistently outperforms others, highlighting the importance of tailoring harness selection to specific problems rather than relying on a universal approach. Additionally, the study introduces an adaptive-allocation strategy that reallocates resources based on early discovery performance, significantly improving outcomes compared to traditional fixed harness methods.
No single discovery harness reliably outperforms others, suggesting that harness selection should be treated as a hyperparameter tailored to specific tasks.
Autonomous discovery systems such as OpenEvolve and TTT-Discover are often used as general-purpose harnesses. However, in practice these are composite systems combining several design choices about archives, parent selection, exploration, and budget allocation into a single recipe. Because discovery runs are expensive and inherently stochastic, existing harnesses are often compared using too few independent trials to distinguish key methodological improvements from run-to-run variance. We systematically decompose OpenEvolve-style evolutionary search and the TTT-Discover search harness into its constituent components and systematically evaluate 30 budget-matched harnesses across 12 model-problem pairs using more than 3.1 million LLM rollouts and repeated-trial statistical analysis. Our results show that discovery harnesses have a generalization problem: No fixed harness is reliably superior across the evaluated model-problem pairs, and variants of OpenEvolve generally underperform simpler alternatives. Thus, harness choice is better viewed as a hyperparameter rather than as a universal recipe, and should be tailored to the specific problem and underlying model. We also find that early discovery progress predicts final performance, and use this property to present a budget-matched adaptive-allocation experiment that starts multiple harnesses, prunes weak partial runs, and reallocates compute to stronger survivors, outperforming both commitment to a randomly sampled fixed harness and a non-adaptive harness ensemble. Together, these results motivate shifting from fixed harness selection to online adaptation guided by early performance. We release all run pools including baseline null distributions for every model-problem pair as reusable statistical infrastructure against for future harness proposals.
BAIR