analiza aiqbench
This study challenges the validity of current evaluation protocols for automatic harness evolution in LLM agents, arguing that shared benchmarks between search and testing introduce significant overfitting risks. By comparing evolved harnesses against simple test-time scaling baselines with matched budgets, the research finds limited generalization gains, suggesting that apparent improvements may stem from additional search rather than superior design. These findings call for fairer evaluation standards to accurately assess the true efficacy of agent optimization techniques.
Rethinking the Evaluation of Harness Evolution for Agents
arXiv:2607.12227v2 Announce Type: replace Abstract: We revisit the evaluation of automatic harness evolution for LLM agents. Existing harness evolution methods use unit test cases to search for harness configurations and then report final performance on the same public benchmark. This protocol raises two fundamental concerns. First, harness evolution is itself an iterative search procedure that repeatedly evaluates and revises candidate harnesses using task feedback. As in agentic test-time scaling, it should therefore be compared with simple task-level search baselines under matched feedback and inference budgets to determine whether its gains arise from improved harness design or from additional search alone. Second, because the search and the final evaluation share the same benchmark, the reported gains risk overfitting to that specific task set. To address these concerns, we conduct an extensive evaluation comparing harness evolution with simple test-time scaling and discovery baselines under comparable feedback and inference budgets, and also evaluate evolved harnesses on held-out tasks to assess whether the discovered improvements generalize. Experiments on Terminal-Bench 2.1 with GPT-5.4 and Claude Opus 4.6 show that automatic harness evolution does not consistently outperform simple test-time scaling methods and exhibits limited generalization. Our results raise important questions about the effectiveness of automatic harness evolution and highlight the need for fairer evaluation protocols and benchmarks for automatic harness design. Our code is available at https://github.com/rethinking-harness-evolution.
Abstract:We revisit the evaluation of automatic harness evolution for LLM agents. Existing harness evolution methods use unit test cases to search for harness configurations and then report final performance on the same public benchmark. This protocol raises two fundamental concerns. First, harness evolution is itself an iterative search procedure that repeatedly evaluates and revises candidate harnesses using task feedback. As in agentic test-time scaling, it should therefore be compared with simple task-level search baselines under matched feedback and inference budgets to determine whether its gains arise from improved harness design or from additional search alone. Second, because the search and the final evaluation share the same benchmark, the reported gains risk overfitting to that specific task set. To address these concerns, we conduct an extensive evaluation comparing harness evolution with simple test-time scaling and discovery baselines under comparable feedback and inference budgets, and also evaluate evolved harnesses on held-out tasks to assess whether the discovered improvements generalize. Experiments on Terminal-Bench 2.1 with GPT-5.4 and Claude Opus 4.6 show that automatic harness evolution does not consistently outperform simple test-time scaling methods and exhibits limited generalization. Our results raise important questions about the effectiveness of automatic harness evolution and highlight the need for fairer evaluation protocols and benchmarks for automatic harness design. Our code is available at this https URL.
Artificial Intelligence (cs.AI)
arXiv:2607.12227 [cs.AI]
(or arXiv:2607.12227v2 [cs.AI] for this version)
https://doi.org/10.48550/arXiv.2607.12227 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 310 cuvinte
Subiecte conexe
EVOHARNESSBENCH: Can Your Agents Keep Pace with an Evolving Harness?
arXiv · 7 sept. 2026
Aspire: Can Models Self-Evolve from Vague Goals?
arXiv · 1 sept. 2026
Don't Blame the Large Language Model: How Agent Harness Evolution Shapes Coding Agent Quality
arXiv · 23 iul. 2026
HarnessDev: Can LLMs Create and Evolve Their Own Agent Harness?
arXiv · 2 sept. 2026