AI Insight
This study introduces Environment Duels, a benchmark that evaluates large language models (LLMs) both as creators of interactive problem environments and as solvers of those problems. Testing nine frontier LLMs revealed that solving ability and authoring ability are distinct skills, with smaller models sometimes creating harder challenges and more useful hints than larger ones. The research found that pairwise interactions between models produce transitive rankings, convergent designs, and that a model's performance improvement from training correlates with how well it uses hints in environments.
Why it matters
This framework provides insight into how LLMs influence each other's development through data sharing and training, which is increasingly relevant as models learn from content generated by other AI systems. The findings could inform more effective training strategies and help predict how models will perform in ecosystems where they interact with content from multiple AI sources.
Understand the Science
⚠️ Preprint – Noch nicht peer-reviewed
Dieser Artikel wurde noch nicht von unabhängigen Experten begutachtet. Die Ergebnisse sind vorläufig und sollten mit Vorsicht interpretiert werden.
Abstract: Through a tangle of real-world interactions, from synthetic training pipelines to data shared on the open web, frontier LLMs absorb what their predecessors and competitors create. How well does each model play its role in this ecosystem, as a creator and as a solver? We introduce Environment Duels, a dynamic agentic benchmark that evaluates both capabilities on interactive environments. Each model writes environments, each with a privileged hint, aiming for tasks that capable solvers fail without the hint and pass with it. Across nine frontier LLMs, solving and authoring prove to be distinct skills: the strongest solvers are not the strongest authors, while smaller models learn more from how others played and write the hardest environments and the most useful hints. Beyond the scores, the pairwise interactions form a microcosm of how models affect one another. Head-to-head outcomes are transitive and so yield a stable ranking; independent authors converge on the same designs; and in a reinforcement-learning study with Qwen3.8-27B, the learner’s hint lift on an author’s environments tracks how much training on those environments improves it on held-out reasoning and agentic benchmarks, whether the environments are its own or another model’s. Each of these patterns has a counterpart in how models shape one another outside the duel. By staging these interactions under controlled conditions, Environment Duels offers a window onto how LLMs teach, challenge and learn from one another. Code is available at https://github.com/LLM-Duels/EnvDuels.
Source: Environment Duels: A Microcosm of LLM Co-Evolution