# Researchers alternate agent sessions on machine-learning tasks > Humanize MLE Agents alternates two agents, resetting session context while preserving a shared workspace. - Canonical URL: https://newsverum.com/en/news/multi-agent-session-alternation-tests-machine-learning-engineering - Publisher: Newsverum - Section: Technology - Country: China - Published: 2026-10-13T00:51:41+00:00 - Updated: 2026-10-13T00:51:41+00:00 - Event / data period: 2026-10-09 - Other editions: [Türkçe](https://newsverum.com/tr/news/multi-agent-session-alternation-tests-machine-learning-engineering.md), [العربية](https://newsverum.com/ar/news/multi-agent-session-alternation-tests-machine-learning-engineering.md), [Français](https://newsverum.com/fr/news/multi-agent-session-alternation-tests-machine-learning-engineering.md), [Español](https://newsverum.com/es/news/multi-agent-session-alternation-tests-machine-learning-engineering.md), [Deutsch](https://newsverum.com/de/news/multi-agent-session-alternation-tests-machine-learning-engineering.md), [Português](https://newsverum.com/pt/news/multi-agent-session-alternation-tests-machine-learning-engineering.md) Researchers affiliated with Tsinghua, Zhejiang and NVIDIA presented Humanize MLE Agents, a method that alternates two agents’ sessions on long machine-learning engineering tasks. Each session stops after a capped number of experiments; context resets between sessions while code, models and prior results remain in a shared workspace. Across 75 Kaggle tasks in MLE-bench, the authors report a 78.2% any-medal rate and say the workflow surpassed the strongest open-source comparison with a smaller time budget. This is a result for a specific competition set and agent configurations; it does not guarantee the same gain across model versions or real engineering projects. ## Data ### Measure 75 Kaggle tasks; MLE-bench evaluation Methodology: Values come from the paper's stated benchmark experiments; they are not field measurements or general performance guarantees. | Measure | Reported value | Experimental scope | |---|---|---| | Any-medal rate on MLE-bench Kaggle tasks | 78.2 | 75 Kaggle tasks; MLE-bench evaluation | Sources: · Download data: [JSON](https://newsverum.com/api/v1/articles/multi-agent-session-alternation-tests-machine-learning-engineering/data/reported-result-1?lang=en&format=json) · [CSV](https://newsverum.com/api/v1/articles/multi-agent-session-alternation-tests-machine-learning-engineering/data/reported-result-1?lang=en&format=csv) ### Measure Kaggle task set reported in the MLE-bench evaluation Methodology: Values come from the paper's stated benchmark experiments; they are not field measurements or general performance guarantees. | Measure | Reported value | Experimental scope | |---|---|---| | Number of MLE-bench Kaggle tasks | 75 | Kaggle task set reported in the MLE-bench evaluation | Sources: · Download data: [JSON](https://newsverum.com/api/v1/articles/multi-agent-session-alternation-tests-machine-learning-engineering/data/result-75?lang=en&format=json) · [CSV](https://newsverum.com/api/v1/articles/multi-agent-session-alternation-tests-machine-learning-engineering/data/result-75?lang=en&format=csv) ## Sources - [arXiv preprint by Tsinghua University, Zhejiang University and collaborating researchers]() (Source checked: 2026-10-13; Published: 2026-10-09) --- Cite the canonical URL above. [Editorial standards and corrections](https://newsverum.com/en/editorial)