Abstract
Language model agents can be improved by updating their model weights or refining the harness that guides task execution. These components are coupled: weight updates change how the model uses the harness, while harness updates change the trajectories used for training. We propose VACE, Validation-Gated Alternating CoEvolution, which alternates agentic reinforcement learning with trajectory-driven harness refinement. After each RL stage, VACE reuses the collected trajectories to propose a harness revision and evaluates the incumbent and candidate with the updated model held fixed. The candidate guides subsequent training only if it improves validation performance. With Qwen3.5-9B, VACE achieves 45.26% test accuracy on OfficeQA and a mean partial-credit score of 75.19% on AutomationBench, exceeding weight-only RL by 6.43 and 9.09 percentage points and ungated alternation by 4.59 and 6.95 points, respectively. Across 44 harness proposals, 17 reduce validation performance at the updated checkpoint and are rejected before subsequent RL training, highlighting the importance of validation gating.
Keywords
Publication details
- Journal
- Not available
- Open access
- Green open access
Cite this article
APA 7
Qi, J., He, Y., Liu, J., Huang, Q., Hu, S., Dang, Z., Chen, G., Yang, R., Jiang, W., Liu, Y., Lyu, T., & Li, F. (2026). VACE: Validation-Gated Alternating Co-Evolution of Agent Models and Harnesses. https://omanscience.com/en/articles/vace-validation-gated-alternating-co-evolution-of-agent-models-and-harnesses
MLA 9
Qi, Jiexing, et al. "VACE: Validation-Gated Alternating Co-Evolution of Agent Models and Harnesses." https://omanscience.com/en/articles/vace-validation-gated-alternating-co-evolution-of-agent-models-and-harnesses.
Chicago (author–date)
Qi, Jiexing, Yu He, Jun Liu, Qichen Huang, Shaohua Hu, Zhan Dang, Guohua Chen, Rui Yang, Wen Jiang, Yang Liu, Tao Lyu, and Fangming Li. 2026. "VACE: Validation-Gated Alternating Co-Evolution of Agent Models and Harnesses." https://omanscience.com/en/articles/vace-validation-gated-alternating-co-evolution-of-agent-models-and-harnesses.
Harvard
Qi, J., He, Y., Liu, J., Huang, Q., Hu, S., Dang, Z., Chen, G., Yang, R., Jiang, W., Liu, Y., Lyu, T. and Li, F. (2026) 'VACE: Validation-Gated Alternating Co-Evolution of Agent Models and Harnesses', Available at: https://omanscience.com/en/articles/vace-validation-gated-alternating-co-evolution-of-agent-models-and-harnesses.
Vancouver
Qi J, He Y, Liu J, Huang Q, Hu S, Dang Z, et al. VACE: Validation-Gated Alternating Co-Evolution of Agent Models and Harnesses. https://omanscience.com/en/articles/vace-validation-gated-alternating-co-evolution-of-agent-models-and-harnesses
IEEE
J. Qi, Y. He, J. Liu, Q. Huang, S. Hu, Z. Dang, G. Chen, R. Yang, W. Jiang, Y. Liu, T. Lyu, and F. Li, "VACE: Validation-Gated Alternating Co-Evolution of Agent Models and Harnesses," https://omanscience.com/en/articles/vace-validation-gated-alternating-co-evolution-of-agent-models-and-harnesses.