{"slug": "cross-entropy-games-and-frost-training", "title": "Cross-Entropy Games and Frost Training", "summary": "Researchers introduced Frost Training, a method that improves Monte Carlo-based policy optimization for Cross-Entropy Games by exploiting the gradient of the reward function in embedding space. The technique, validated using GRPO training for maximum-likelihood infilling, enables models to generate higher-scoring outputs at increased speed in best-of-k settings. This marks the first demonstration that the gradient signal used in the Greedy Coordinate Gradient jailbreaking technique can also boost model training.", "body_md": "# Computer Science > Artificial Intelligence\n\n[Submitted on 26 May 2026]\n\n# Title:Cross-Entropy Games and Frost Training\n\n[View PDF](/pdf/2605.27701)\n\n[HTML (experimental)](https://arxiv.org/html/2605.27701v1)\n\nAbstract:We present Frost Training, a method for improving Monte Carlo-based policy optimization for a large family of LLM-as-a-judge tasks called Cross-Entropy Games. The key idea is to exploit the gradient of the reward function in embedding space. This signal is used in the Greedy Coordinate Gradient (GCG) jailbreaking technique; we demonstrate for the first time that it can also be used to boost model training. We validate our method using GRPO training for maximum-likelihood infilling. Frost Training improves the model's ability to generate high-scoring outputs, reaching higher maximum scores in a best-of-k setting, and does so at an increased speed.\n\n### References & Citations\n\nLoading...\n\n# Bibliographic and Citation Tools\n\nBibliographic Explorer\n\n*(*[What is the Explorer?](https://info.arxiv.org/labs/showcase.html#arxiv-bibliographic-explorer))\nConnected Papers\n\n*(*[What is Connected Papers?](https://www.connectedpapers.com/about))\nLitmaps\n\n*(*[What is Litmaps?](https://www.litmaps.co/))\nscite Smart Citations\n\n*(*[What are Smart Citations?](https://www.scite.ai/))# Code, Data and Media Associated with this Article\n\nalphaXiv\n\n*(*[What is alphaXiv?](https://alphaxiv.org/))\nCatalyzeX Code Finder for Papers\n\n*(*[What is CatalyzeX?](https://www.catalyzex.com))\nDagsHub\n\n*(*[What is DagsHub?](https://dagshub.com/))\nGotit.pub\n\n*(*[What is GotitPub?](http://gotit.pub/faq))\nHugging Face\n\n*(*[What is Huggingface?](https://huggingface.co/huggingface))\nScienceCast\n\n*(*[What is ScienceCast?](https://sciencecast.org/welcome))# Demos\n\n# Recommenders and Search Tools\n\nInfluence Flower\n\n*(*[What are Influence Flowers?](https://influencemap.cmlab.dev/))\nCORE Recommender\n\n*(*[What is CORE?](https://core.ac.uk/services/recommender))# arXivLabs: experimental projects with community collaborators\n\narXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.\n\nBoth 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.\n\nHave an idea for a project that will add value for arXiv's community? [ Learn more about arXivLabs](https://info.arxiv.org/labs/index.html).", "url": "https://wpnews.pro/news/cross-entropy-games-and-frost-training", "canonical_source": "https://arxiv.org/abs/2605.27701", "published_at": "2026-05-28 04:00:00+00:00", "updated_at": "2026-05-28 04:32:35.732082+00:00", "lang": "en", "topics": ["artificial-intelligence", "machine-learning", "large-language-models", "ai-research"], "entities": ["Frost Training", "Cross-Entropy Games", "GRPO", "Greedy Coordinate Gradient", "GCG", "LLM-as-a-judge"], "alternates": {"html": "https://wpnews.pro/news/cross-entropy-games-and-frost-training", "markdown": "https://wpnews.pro/news/cross-entropy-games-and-frost-training.md", "text": "https://wpnews.pro/news/cross-entropy-games-and-frost-training.txt", "jsonld": "https://wpnews.pro/news/cross-entropy-games-and-frost-training.jsonld"}}